Table of contents

Artificial intelligence(AI) and Big Data make it possible for machines to learn from experience, adjust to new inputs and perform human-like work. Most AI examples that you hear about today – from chess-playing computers to self-driving cars – rely heavily on deep learning and natural language processing. Computers can be trained to accomplish specific tasks by processing large amounts of data and recognizing patterns in the data. It is now changing our world greatly, its development depends heavily on the artificial intelligence theory and technology progress, so its related researches is now becoming hot all over the world.

ICCBDAI 2020 brought together an international community of researchers and practitioners in the field of the related domain to discuss the latest advancements of the discipline, shape its future directions, and promote its diffusion among the scientific community at large in Changsha, China.

List of ICCBDAI 2020, Group Photo, Scene Photos and this titles are available in this pdf.

All papers published in this volume of Journal of Physics: Conference Series have been peer reviewed through processes administered by the Editors. Reviews were conducted by expert referees to the professional and scientific standards expected of a proceedings journal published by IOP Publishing.

• Type of peer review: Single-Blind. The members of Reviewer Team reviewed papers by creativity, scientific value, the relevance to the topics, paper format requirements, and English proofread.

• Conference submission management system: Authors submit papers via mail system

• Number of submissions received: 368

• Number of submissions sent for review: 360

• Number of submissions accepted: 209

• Acceptance Rate (Number of Submissions Accepted / Number of Submissions Received X 100): 56%

• Average number of reviews per paper: 2

• Total number of reviewers involved: 40

Any additional info on review process: 1. Single-blind peer review (by one of the editors) with editorial entry check (by the corresponding editor) and post-acceptance language and style check (by an invited linguist).

2. The review process for each paper took around 1 to 2 month/s. In the case of contradict decision of acceptance (accepted of reject)/suggestion for publication between two reviewers of the paper, the final selection was made based on either weighted all comments or initiated the third reviewer for new weighted selection. And the core of the review team is comprised from educators and researchers from universities and research institutions at home and at abroad.

Contact person for queries: Miss Xiaoai Zhao

Email: ywdjsj@gmail.com

Spectral clustering is one of the most popular clustering approaches. Despite its good performance and strong theoretical supports, it is limited to high complexity of the graph Laplacian similarity matrix construction and eigen-decomposition problems. Recently, deep learning has been successfully adopted in graph representation. In the paper, we jointly learn the manifold graph construction and non-linear low-dimension mapping of the graph. In addition, we theoretically proved that our model according with spectral clustering theory. Meanwhile, we use the proposed non-linear coders as the building blocks to formulate a deep structure to further refine features of layer wise fashion. Extensive experiments on clustering tasks demonstrate that our method performs well in terms of both clustering accuracy and normalized mutual information( NMI )

With the fish of crisped grass carp as the research object, this paper analyzes the hardness-hyperspectral data of crisped grass carp with a texture analyzer and a visible - near infrared hyperspectral imaging system (400-1100nm), and then proposes a method to quickly detect the hardness of crisped grass carp. In this study, the hardness of 15 fish crisped grass carp was analyzed with a texture analyzer, and it was found that there was a difference in hardness between different fish with a significance P<0.05.Four spectral bands affecting the hardness of crisped grass carp were found by using the random forest algorithm, which indicated that there was a connection between hardness and hyperspectrum. Four pretreatment methods were used, among which SNV pretreatment method had the best effect, R²was 0.80. PLS, SVM and BP-NN prediction models were established respectively, and R²and RMSEP of the models were used as evaluation indexes to compare the advantages and disadvantages. The results are as follows: R²is 0.8033, 0.8564, 0.8814 and RMSEP are 3.6417, 2.6891 and 2.6128 respectively, among which BP-NN prediction model R²is 0.8814 and RMSEP is 2.6128, which has the most significant effect. Therefore, BP-NN prediction model is more suitable for nondestructive testing of crisped grass carp hardness.

In recent years, the soaring development of CNN has facilitated the maturity of the Computer Vision Algorithm. This paper will briefly introduce some representative Target Detection Algorithm, and systematically analyze the underlying problems, the modified methods, and the prospective direction of the algorithm in accordance with its merits and demerits. It is generally divided into a single-stage detection model and a double-stage detection model in terms of whether candidate areas need to be extracted during target detection for further tasks. Featured with scale in the double-stage detection model, the algorithm is divided into single-scale detection and multi-scale detection based on whether it can appropriately integrate with a network structure, which enhances the accuracy of the network model towards small targets. Meanwhile, it can also be divided into anchor-base and anchor-free in a single-stage detection model on the basis of the anchor bolt. A predictable development of the Target Detection Algorithm will show to us in the future.

Fodder, fish manure, and pond sludge will seriously affect the turbidity of the water body in aquaculture. How to quickly and online judge the turbidity of the water body is very important for realizing efficient, low-cost, and accurate control of aquaculture. In view of the shortcomings of traditional detection methods, a transfer learning method based on ResNet deep learning network model is proposed to realize water body turbidity classification, and two transfer learning methods of parameter partially frozen and completely unfrozen are designed based on ResNet18.Subsequently, the turbidity data set of aquaculture water quality was constructed, and the data set was enhanced by image cropping, image flipping, random scaling, and other methods. The experimental results show that when all parameters are not frozen, the transfer learning method can achieve the best class effect, and the accuracy rate is 0.9686, which can provide an effective method for online detection of aquaculture water body turbidity.

Social unrest is endemic in many societies. It is a vital task to predict social unrest events be-cause these events can lead to societal changes and endanger public security. We combine two data sources, using social media data as the historical texts and news media data as the ground truth of unrest events. We propose a temporal dual graph convolutional network (TDGCN), which extracts the contextual semantic information and the communication relationship between social network users from the historical texts and constructs two dynamic graphs to capture the implied semantic and temporal features. The TDGCN can predict the occurrence of unrest events. Experimental results on a specific data set of unrest show that the proposed method has better performance than other state-of-the-art social unrest event predictions.

The traditional traffic classification task relies on feature engineering by experts with specialized knowledge. With the advent of representation-learn-based traffic classification techniques, machines learned to extract features from traffic data and classify them. In this paper, we research traffic classification technology based on representational learning and import auto-machine learning technology to solve the problems of network architecture design and parameter tuning. We also design a befitting reward function for the network architecture model. The experimental results on the USTC-TF2016 dataset and USTC-TF2016-PLUS shows that the network architecture generated by auto-machine learning technology has better training performance and classification accuracy than a traditional neural network.

In view of the complexity of traditional document similarity calculation methods and the problem of large error, a document similarity calculation and detection method based on deep learning is proposed. The effective subtree matching method is used to calculate the similarity of document feature sequences, and the frequency of feature items is obtained. In order to reduce the complexity of document similarity detection, a deep learning method is used to classify documents. The keywords in the document are extracted, and the similarity calculation and detection results are obtained by calculating the similarity between keywords. The experimental results show that the calculation error of the proposed method is low and the results are reliable, which fully shows that the method is effective and feasible.

The whale optimization algorithm (WOA) has been widely used in different applications. It has simple control parameters and novel population updating mechanism. However, there is few theoretical analysis of WOA and the convergence property of WOA is ambiguity. This paper analyzes the convergence property of WOA by using the Markov chain of the stochastic process theory. The Markov chain model of the WOA algorithm is established. The one step transition probabilities and convergence properties of different population updating mechanisms in WOA are given. It's proved that the convergence property of WOA is determined by its shrinking encircling mechanism. Finally, three algorithms with different population updating mechanisms are tested with thirteen benchmark functions on accuracy and convergence speed. The simulation results on benchmark functions verify the validity of the theoretical analysis of WOA.

In order to improve the imaging effect of geographic images in a complex weather environment. This paper designs a geographic image defogging algorithm by using the generation countermeasure network to improve the clarity of the image. Firstly, the imaging mode and degradation process of the foggy image is studied in this paper, and the atmospheric scattering model is established to simulate the degradation of the image. Secondly, according to the data characteristics of image defogging, an optimized image defogging model is generated based on Conditional Generation Adversarial Nets (CGAN), so that the image output by the model is closer to the real image. The experimental results show that the image defogging model designed in this paper has good contrast and color richness, which can effectively solve the problem of image degradation in fog. Moreover, the reduction of paired data sets will reduce the measurement index of the geographic image defogging model. This paper provides a reference for the study of geographic image defogging.

With the development of artificial intelligence, using machine learning methods to build evaluation models has attracted more and more attention. However, training anevaluation model often needs a lot of labeling samples annotated by experts. It is very difficult to get enough labeling data with a limited group of experts. This paper proposes a method to learn the evaluation model with limited information from experts. This method has two stages. In the first stage, we build a large training set with ordinary people by comparing every two samples. After that, we train a Siamese Network with the paired comparison data set to get a score for each sample. In the second stage, we map the scores to evaluation grades with the help of experts. In the experiments, we use the UCI wine quality data set to evaluate our method. Experimental results demonstrate that we get a basically equivalent accuracy (0.5% decrease) with only1.45% samples labeled byexperts.

With the development and popularization of artificial intelligence and machine vision, human posture recognition has become one of the most important research directions in the field of machine vision, but its application research in dance movement recognition is in its infancy, especially in the field of Chinese classical dance. In view of this situation, the application of artificial intelligence in basic hand position recognition of Chinese classical dance. Including analysis of basic hand position of Chinese classical dance, comparison of human posture recognition models, recognition and comparison of basic hand position of Chinese classical dance on NVIDIA embedded devices using TensorFlow deep learning framework, and combination of traditional culture and artificial intelligence technology.

In this paper, an intrusion detection model (ACGANs-CNN) method based on GAN and CNN fusion is proposed for the reasons that unknown attack sample data cannot be provided in training samples, the number of training samples is limited, and known attack sample types account for less such small sample data. The model converts network traffic data into grayscale images, generates the same proportion of attack samples by generating the counter network, ensures the uniform distribution of attack samples in the training set, and introduces the gradient penalty function to improve the stability of the training model. Secondly, CNN is used to better extract sample features. In order to prevent overfitting, the nonlinear activation function Relu and Dropout method are introduced. At the same time, the convergence speed of the model is accelerated, and the detection efficiency of the model is improved. Attention is introduced to highlight the key features and to classify samples based on these key features. In this paper, the KDDCUP99 data set is used for model evaluation. Experimental results show that this algorithm (ACGANs-CNN) has stronger model training stability, higher quality of generated fake samples, and better feature extraction effect in small sample data. Its detection rate and accuracy of attack types are significantly higher than that of traditional machine learning algorithms such as SVM, KNN, RF, and other CNN models.

EIT (Electrical Impedance Tomography) is a non-invasive dynamic image detection technology that can well reflect the distribution of objects in a uniform field. But the inverse problem of EIT is a non-linear, highly ill-conditioned, and ill-posed problem. Images reconstructed by traditional iterative algorithms and non-iterative algorithms have too many uneliminated artifacts and low spatial resolution. This paper proposes a deep neural network containing a multi-layer fully connected network so that predicting the conductivity distribution of different samples through the excellent nonlinear fitting ability of the neural network. Compared with the CG(Conjugate Gradient) algorithm and TR ( Tikhonov Regularization) algorithm, the quality of images is improved and the noise is reduced, but the generalization ability and prediction accuracy of the network need to be further improved.

There are many defects in current classroom teaching evaluation, which is not conducive to the implementation of quality education. Therefore, the AI-based classroom teaching evaluation technology is proposed, using artificial intelligence theory, from the perspective of serving quality education, analyzing the problems in teaching evaluation, exploring effective classroom teaching evaluation, and ultimately promoting the development o f students. Starting from the theory of artificial intelligence, it discusses the necessity of constructing a multi-evaluation system for the teaching process, which improves students' language application ability and humanistic quality, and applies artificial intelligence theory to construct teaching from the aspects of evaluation methods, evaluation methods, evaluation objects, and evaluation subjects. Evaluation System. Experimental results show that this technology can accurately evaluate teaching effects and improve teaching quality

The development of embedded technology directly affects the development of Internet of Things, and the development of artificial intelligence brings great convenience to people's life. Based on this, the implementation method of artificial intelligence algorithm in embedded system is studied. Based on embedded system hardware configuration and the embedded system software running algorithm is optimized, simplifying the operation steps of the embedded system, improve the effect of embedded system, to strengthen the research of embedded technology, and increase investment in manpower and material resources in embedded system research, learning western advanced technology, and the embedded technology is applied in the Internet of things, to improve the effect of artificial intelligence algorithm in the application of the embedded system. The experimental results show that the artificial intelligence algorithm is applied to the embedded system to improve the system performance effectively.

In order to improve the security of drilling progress and reduce the reliability impact which caused by the collision between drill string and well wall, the finite element analysis software ADINA is used to establish a finite element model to simulate the change of drill string stress after the collision between drill string and wellbore under different factors. The results show that the existence of drilling fluid with higher viscosity, lower revolution speed and autorotation speed of drill string and larger annular space is conducive to reduce the stress value caused by collision between drill string and wellbore and improve the security and stability of the drill string.

Artificial intelligence affect the management of internet content products. This paper defines the internet content products, figures out the model of internet content production process, proposes a profit model of Internet content operation. It also summarizes the impact of artificial intelligence on the operation of Internet content products.

Driven by big data and artificial intelligence technology, the core business of advertising is gradually becoming intelligent, from consumer insight to advertising information delivery and then to advertising content production. With the continuous progress of cognitive computing technology, it can be predicted that in the future, intelligent advertising will continue to step forward along the trend of "humanization" in line with human needs.

China has a wide range of engineering students sources and high quality. but because of the low level of internationalization of education, it is unable to meet the needs of international talents in economic globalization. Artificial intelligence has stepped into the education industry, bringing opportunities for foreign language teaching to cultivate engineering talents with an international perspective. In the era of artificial intelligence, the construction of foreign language curriculum system in engineering majors should set higher-level curriculum teaching objectives, develop innovative teaching system and improve the teaching assessment of high-challenge courses guided by OBE education concept in order to improve the quality of personnel training.

Based on the operator's competency index matrix, a two-layer mathematical model with the goal of maximizing the total competency of employees' workstations is designed. The upper layer uses the Hungarian Algorithm to achieve the configuration of "strong employees", and the lower layer uses the Hybrid Genetic Simulated Annealing Algorithm to achieve the configuration of "weak employees". The results show that: compared with the traditional undifferentiated overall configuration. The double-level configuration based on the double-level mathematical model increases the employee competency difference ratio by 5.71%, which effectively embodies the idea of "strong leads the weak".

A new emotion recognition system based on speech is constructed to improve the ability of recognizing negative emotions. Multi-dimensional acoustic characteristics were tested and among them, short-term energy and Mel-frequency cepstral coefficients (MFCC) were selected to be used as parameters for recognition. The system consists two modes: single recognition and group recognition. Single recognition adopts BP neural network model based on MFCC, while group recognition adds support vector machine model based on short-term energy on the basis of single recognition which the group recognition rate of 20 speech can reach 97%. With the increase of the number of speech in each group, the recognition accuracy of negative emotion tends to 100%.

The current rapid development of the remote sensing satellite industry provides a large amount of image data for ship classification tasks. Aiming at the problem of insufficient feature extraction of single source image, this paper designs a lightweight ship classification model based on the fusion of panchromatic image and multispectral image of pseudo Siamese network to extract image features more fully. First, establish a multi-source remote sensing image ship target classification dataset MPFS (MS and PAN Ship image Fusion Classification Dataset); secondly, send panchromatic images and multispectral images to the network through different convolutional layers, thendesign a multi-level feature extraction network for panchromatic images and an adaptive feature extraction network for spectral imagesrespectively; finally, concatenate the features along the channel dimension and send them to the classification network.

Response selection in retrieval-based chatbot aims to find the most relevant response in a candidate repository given the conversation context. A key technique to this task lies in how to measure the matching degree between conversation context and response at rich semantic information. In this paper, we propose a hierarchical residual matching network (HRMN) to fully extract and make use of the rich semantic information in the conversation history and response for themulti-turn response selection task. We empirically verify HRMN on two benchmark data sets and compare against advanced approaches. Evaluation results demonstrate that HRMN outperforms strong baselines and has a distinct improvement in response selection.

Benefiting from its simplicity and efficiency, particle swarm optimization (PSO) algorithm has shown great performance on various problems. However, for different optimization problems or different search areas, it is still difficult to achieve a satisfying trade-off between exploration and exploitation. On the basis of canonical PSO algorithm, a variety of improved algorithms have been proposed, which have different capabilities of exploitation and exploration, and each of them performs effective in some problems. This paper proposes a particle swarm optimization with multiple adaptive sub-swarms (PSOMAS). It uses multiple subswarms strategy, in which each sub-swarm is evolved by different algorithms, and an adaptive strategy is also used to reduce the consumption of computing resources. A comprehensive experimental study is conducted on 30 benchmark functions, to compare with several well-known variants of PSO algorithms. The results show that PSOMAS with RT=100 could obtain a better overall performance than all others. Moreover, PSOMAS could find high-quality solution in different problems by varying the value of RT.

The bi-frequency (high- and low) synthetic aperture radar (SAR) images cannot be directly compared due to their distinct statistical properties. To diminish their statistical difference, we manage to translate the bi-frequency SAR images into one another. Therefore, we propose a cycle-consistent conditional adversarial network to achieve the goal. The cycle-consistency criteria in the Cycle GAN and the conditional generation adversarial networks in the Pix2Pix are integrated to construct the cycle-consistent conditional adversarial network. Experiments on Ku-band and P-band SAR images validate that our method outperforms Cycle GAN and Pix2Pix.

The existing hippocampal modeling approaches rarely span the wide functionality range from processing raw sensory signals to planning and action. This paper presents a goal-directed navigation system consisting of two planning strategies. The first one is a biologically inspired neural planning and navigation model that is related to learned representations of place and HD cells. It is responsible for generating spatial trajectories leading to the neighboring area of the target. The place and HD cells are trained unsupervisedly from visual images using a modified slow feature analysis (SFA) algorithm. To interpret their functional role in navigation, a planning network is trained to predict the neural activities of place and HD cell representations given selected action signals. Recursive prediction and optimization of the action signals generate goal-directed activation sequences, in which the continuous states and action spaces are represented by the population of place-, HD- and motor neuron activities. Furthermore, a second planning strategy relying on visual recognition is proposed and performs target-driven reaching on a local scale for finer accuracy. Experimental results show the effectiveness of the proposed system.

There are a large number of unstructured texts without data cleaning in the field of electric power. It is expensive to rely on manual ways to process the large amount of text data. In order to reduce the workload of data cleaning, we propose an intelligent method of automatic revision of power defect logs in this paper by adopting natural language processing technologies. We utilize entity recognition technology to recognize electrical equipment words on the text and utilize word similarity calculation to find out words with similar meaning in the standard vocabulary, which is the main process to revise Abnormal text. With the outstanding performance of entity recognition, the workload of data cleaning is reduced approximately 70% through our proposed method, which greatly improves the efficiency of unstructured data processing.

Thousand-kernel weight is an important agronomic parameter of rapeseed. In order to quickly realize the grain count, shorten the measurement cycle of the thousand-grain weight, in this paper, we propose a grain counting method based on image detection, by testing 300 rapeseed images, and the results showed that the detection rate reached 89.33%.

Knowledge Graph provides an effective scheme for unstructured knowledge on the Internet. but, to a large extent, the lack of information in the knowledge graph restricts the application of the knowledge graph. Therefore, knowledge graph completion has been an attractive research area. In recent years, many researchers apply graph convolutional neural networks to knowledge graph embedding so as to solve this problem, including R-GCN, SACN et al. The current state-of-the-art SACN uses graph convolutional neural as an encoder to make more accurate embeddings of graph nodes, and uses a convolutional network as a decoder, leading a good performance of link prediction task. In this work, we produce a novel graph representation model based on SACN---property graph convolutional network called PGCN. PGCN treats the knowledge graph as a property graph, regarding the initial embedding vector of entities and relations as the property of nodes and edges. What makes the model different is that it introduces the node clustering before convolving the nodes, so that the graph takes into account the importance of neighbors when aggregating the neighbors of nodes. We adopt the probability model Conv-TransE proposed by SACN for the modeling of relation of graph, Conv-TransE takes advantages of ConvE and use probability method which greatly improves the efficiency of the experiment and avoids the construction of the corrupted triplet. We conduct experiments on the standard datasets WN18RR and FB15k-237, and demonstrate that our new model achieves better performance.

Succession command is a command method to change command in the face of uncertain factors or irresistible damage. Alternative command nodes need to complete command transfer to achieve uninterrupted command. Facing the alternative scenarios of multiple command nodes, this article proposes a method for selecting the replacement command nodes, researches the command and control capability requirements, analyzes the command capability of the command organization, extracts the key performance indicators of the command and control system to construct an evaluation index system, and uses the analytic hierarchy process to determine the index rights. Based on the value, a hybrid TOPSIS method and grey relational analysis evaluation model is proposed to evaluate and sort the candidate command nodes. Experiments are carried out through the simulation data of the command system. The experimental results prove that the method can effectively select the replacement command nodes and provide preparation for the subsequent replacement command.

Gestures serve as the best alternative to traditional human-computer interaction (HCI), but there is still a great challenge to apply gestures to practical operations. Faced with the problem of generally low recognition accuracy in dynamic gesture recognition, we propose a fusion network with a slow-fast structure based on an attention mechanism to improve the recognition accuracy of dynamic gestures. The slow pathway acquires the temporal information of the input dynamic gesture, the fast pathway acquires the semantic information of the gesture in the input video, and suppresses the influence of non-gesture regions on the gesture features as much as possible through the attention mechanism, and finally performs the fusion operation according to the strategy of score fusion to obtain the recognition accuracy of the input dynamic gesture. We validate our proposed method on the ChaLearn large-scale gesture challenge gesture dataset IsoGD, and the experimental results are obtained to verify the effectiveness of our proposed structure by comparing it with the previous experimental results.

Aiming at the disadvantage of the SSD algorithm's poor detection effect on small objects in the process of detecting objects, a detection method based on feature fusion and feature enhancement is proposed. Our model combines deep and shallow features to improve the model's ability to detect small objects. Besides, the idea of a channel attention mechanism is introduced into the model structure to increase the feature extraction ability of the network and effectively improve the detection accuracy of the network. Experimental results show that the detection accuracy (mAP) is 2.3% h igher than that of SSD. The detection accuracy (mAP) is significantly improved compared to the current series of excellent object detection algorithms. The algorithm in this paper can detect small objects that are not detected by most SSD algorithms, which improves the average detection accuracy

To reduce the security risks and guide management pressure of "mountain-type" scenic areas with steep and changeable topography, this paper combines the idea of swarm security intelligence and proposes a swarm security intelligent dynamic networking algorithm based on multiple constraints to ensure the swarm safety of the tour group and relieve the management pressure of tour guide through the dynamic group and the self-discovery after leaving the group. Based on multiple constraints such as signal strength, power consumption and fault probability etc, the algorithm optimizes the network topology in real-time, which can effectively reduce the high false-positive rate o f leaving the group due to node failure. Secondly, the algorithm constructs the Spf constrained fitness function and introduces the ant colony optimization algorithm to avoid the situation that the optimal routing path falls into the local optimal solution. Finally, the algorithm converts the adjacency matrix composed of routing table and node connection signals into areachable matrix to judge whether the member has left the group and reduce the false-negative rate of group loss in the case of multi-node connection. Experimental results show that the algorithm has good stability and robustness in mobile wireless adhoc networks

Trad itional machine learning sentiment analysis models are d ifficult to achieve good classification results from small sample data. This paper proposes to merging naive Bayes and causal rule(MNBACR) for small sample data sentiment analysis scenarios. This model is based on the causal analysis theory, and introduce the causal inference algorithm into the field of text sentiment analysis. The causal inference algorithm extracts the causal ru les of Chinese texts, and the causal rules can be used as the features of the naive Bayes algorithm to predict the sentiment polarity of small sample texts. In experiments, the model in this paper is evaluated on financial news datasets which have a small number for sample, and the results show that the proposed method achieves the best performance compared to the existing state-of-the-art models on the small sample data onto sentiment analysis

Solving the problem of pedestrians being occluded by objects is extremely challenging. Using part-level features to describe pedestrian images can provide fine-grained information. However, only paying attention to the local features of body will lack global pedestrian information. And the network consumes time and memory. To solve these problems, we propose a new person re-identification network. The network uses a global contrastivemodule to obtain the features of pedestrians. Through effective use of the pedestrian's global features, as well as the pedestrian's personal information and global contrastive information, the pedestrian can be found in the object occlusion to provide a reliable feature embedding. Our model is tested on Market1501, Duke MTMC-reID, CUHK03 and MSMT17 datasets. The experimental results show that our method is effective in occluded person re-identification

Now is the information age, artificial intelligence rises under the background of this era, but in the field of artificial intelligence, deep learning stands out in machine learning with its excellent learning ability and is rapidly applied to face recognition engineering. However, the success of mainstream deep learning often depends on a lot of training data and training time. This paper proposes to use meta-learning technology to realize face recognition. metalearning is to use the prior knowledge and experience to guide the learning of new tasks and has the ability to learn so that it can avoid the phenomenon of overfitting in the case of only a small number of samples. We regard the data set as a task, that is, meta-train task and meta-test task, using the MAML algorithm [1] fine-tuning gradient model based on prior knowledge to obtain optimization parameters to realize face recognition. Experiments show that deep learning neural networks using meta-learning technology can achieve higher accuracy and rate than ordinary face recognition neural networks. In the experiment, the recognition rate of the face can reach more than 92.6%. It has more intelligence and has made a contribution to the development of face recognition technology.

This essay makes a series of explorations on Vocational school's English teaching mode on the basis of OBE concept and on the age of big data, analyzes its research significance, research contents and main innovations, Therefore, it puts forward the innovation path of Higher Vocational English teaching mode based on OBE concept. The final conclusion is that the h igher vocational English teaching mode based on the OBE concept under the background of big data can promote college English teaching, and can promote vocational schools to be employment-oriented by using big data

Among various natural language process tasks, sentiment analysis has always been a research hotspot. From the initial sentence-level and document-level coarse-grained sentiment analysis to recent fine-grained sentiment analysis on the aspect word level, researchers are committed to applying diverse methods to obtain better sentiment analysis results, ranging from lexicon-based, statistical machine learning methods to deep learning models. In the change of technology, several benchmark datasets that can be used for model performance comparison are gradually yielded. This article summarizes the current research status of aspect-level text sentiment analysis from multiple dimensions such as dataset, mainstream methods, and evaluation indicators, finally, it puts forward the challenges facing and potential research directions from a unique perspective.

The maintenance and inspection of the high-speed railway is the core problem of railway safety. To solve the problems of insufficient network depth and low defect extraction ability, an improved Unet network model was proposed. in view of the relatively small defect area, the attention mechanism is used to effectively suppress the background and highlight the significant features of the defect. The experimental results show that the improved Unet model improves the comprehensive detection accuracy of defect targets by 6% compared with similar segmentation models.

New technologies such as virtual reality, cloud computing, big data, augmented reality, motion-sensing interaction, and artificial intelligence has brought new changes to digital marketing communication talent cultivation. Students should be trained to master the knowledge of artificial intelligence applications and have a broad international vision.

In order to improve the effect of deep learning, this paper puts forward the method of deep learning of Japanese. Before classification, we need to preprocess the word segmentation and delete the termination word in the question, and then use the feature vector to represent it. By learning to store the information in the network, we use heuristic rules to classify the question intention, extract the feature vectors representing different types of questions, and then make statistical analysis on the corpus of actual marked questions, establish a classification system, and bind features based on word packets, Realization intention classification. The experimental results show that the classification accuracy and influence parameters are relatively higher after deep learning, and the automatic classification method of Japanese question intention should be unique.

Deep learning based image completion methods are generally based on three technologies, namely the auto encoder based method, the generation of adversarial networks based method and the recurrent network based method. However, many methods have very single results. In order to obtain the diversity of the completion results, this paper proposes a neural image completion method based on label differentiation, called LD-PICNet (Label Differentiation PICNet). This method can not only generate a complete image with clear and good semantic information, but also actively edit the tags on the generated results to maximize the diversity of the output. Specifically, this paper introduces an auxiliary classifier, which uses a single label of the image ground truth to reconstruct the image, and uses the label to actively increase the difference of the latent vectors during image completion to achieve variability of output. In addition, this paper also introduces a depth-weighted loss function based on information entropy. Different data sets are used to conduct experiments. The model's ability to complete different types of targets is tested. The results show that this method has the ability to generate diversified results, and it has higher clarity than the other methods.

Sleep apnea syndrome is a sleep disease that may lead to sudden death. Long term apnea syndrome can cause chronic cerebral hypoxia, hypertension, cardiovascular and cerebrovascular complications. At present, PSG is the most reliable method for diagnosis. But the diagnosis of PSG is complex and expensive. Electrocardiograph(ECG) and portable medical equipment have been widely used nowadays, which makes the acquisition of ECG signal more and more popular and convenient. In this paper, a convolution neural network based on ECG signal is proposed to predict apnea syndrome, the accuracy and sensitivity of this CNN model for apnea syndrome classification are 94% and 88% respectively. The results show that this method has the advantages of low cost and low complexity.

With the rapid development of intelligent network vehicles, the safety of the vehicle network is taken seriously increasingly, the CAN as the most widely used car on-board network, its security problem has become one of the most important problemsin intelligent made car development, the communication of CAN bus network characteristic and the existing safety problems and analyzes the common means of attack, on the basis of this puts forward the anomaly detection model based on LSTM And ResNet, through the experimental results show that this method CAN effectively detect deception attack.

The deep neural network is a highly expressive model, which plays an extremely critical role in modern artificial intelligence applications. Adversarial samples can change the prediction results of neural networks by imposing imperceptible perturbations on the original images, which brings new challenges to deep learning. In this paper we summarize the methods of generating adversarial samples in recent years, intuitively feel the development of adversarial samples from the time of publication, and briefly classify them from the perspective of algorithm principles. At the same time, for practical applications, the advantages and disadvantages of the algorithm are analyzed and summarized, and the conditions for the algorithm to be suitable for application are proposed: no need to know the specific structure of the algorithm, high quality of the adversarial samples and convenient migration. The analysis points out that among the typical methods, MI-FGSM, ONE-PIXEL, and methods using GAN are more practical and worthy of further study.

In the research of ECG signal identity recognition, most of them adopt the method of feature extraction and recognition model separation, extract the time domain features, transform domain features of the original signal, or combine the features with the cross domain. Then the model is used to complete the recognition and classification. In this paper, an advanced improved convolutional neural network model is proposed, which integrates feature extraction and classification to complete identity recognition. ECG data selected from the ecg-id database and MIT-BIH arrhythmia database are directly sent to the model for automatic sign extraction after hierarchical denoising with wavelet tools and then identified. This method achieves the highest recognition rate of 98.49% on ecg-id database and 99.35% on ECG data of MIT-BIH arrhythmia database. The high reliability of the algorithm and the universality of wireless sensors in mobile devices make this research has high commercial value.

Pneumonia recognition has important research significance in computer-aided diagnosis, and there is a problem of low accuracy for pneumonia recognition. In this paper, an improved network is based on the convolutional neural network AlexNet, the AlexNet_Branch network. The AlexNet_Branch network adds a parallel branch convolutional neural network to AlexNet, and it connects AlexNet and the branch convolutional neural network at the fully connected layer. During training, the same image is simultaneously obtained by AlexNet and the branch convolutional neural network to obtain different feature maps, and then the feature maps are merged together at the fully connected layer to improve the accuracy of recognition. Through design experiments, different AlexNet_Branch networks composed of different layers of branch convolutional neural networks were built, and the network was trained and tested on the chest X-ray image set respectively. The results show that the addition of a branch convolutional neural network greatly improves the accuracy of pneumonia recognition, and the AlexNet_Branch network test accuracy consisting of a 16-layer branch convolutional neural network is 98.01%.

Combining the RPC rational polynomial coefficient model, using WorldView-2 remote sensing images, combined with field surveyed control points, survey area digital elevation model DEM and other data to produce the mountain orthophoto map, combined with the detection points to analyze the accuracy of the results, solve the problem Topographic maps are difficult to measure and are subject to flight restrictions.

The main objective of this paper is to get some basic knowledge about image fusion, the methods used for fusion, and the problems that still exist in image fusion technology, and to draw the attention of the research field to image fusion technology through a combination of theory and practice.

With the popularization of the Internet, the rapid development of intelligent education is promoted. They are a new platform for students to learn, and they can quickly help students improve and consolidate knowledge, as well as unlimited time, place, and space. However, due to a large number of learning resources, how to find the resources you need, and suitable from the massive resources is a problem that needs to be solved in this field. The individual characteristics of each learner are different, so when meeting their own needs, they have different requirements for individualization. The personalized learning resource recommendation system was born, according to the different characteristics of students to recommend corresponding learning resources. Since most of the recommendation systems currently recommend a large number of test questions or a large number of books, but each person's needs are different, this article focuses on the above problems and selects a small piece of content course knowledge points in the learning resources, using course knowledge Come as a recommendation point and solve the above problems by designing a personal personalized learning mechanism. The main research contents of this article are as follows: (1) It is proposed to take curriculum knowledge points as the recommended objects. (2) Use collaborative filtering methods and cognitive diagnosis methods to design individual learning mechanisms. (3) Experiment. Through related experiments, the use of collaborative filtering method and cognitive diagnosis method based on curriculum knowledge points recommendation to confirm whether it is feasible.

The TOR anonymous communication system is an important means to protect network communication security and user privacy, but there are still criminals trying to destroy the confidentiality of the anonymous communication system through some special methods. Aiming at the abuse of the TOR anonymous communication system, this paper proposes a neural network-based anonymous traffic identification method, which uses a one-dimensional convolutional neural network for feature extraction, prediction and classification, and finally integration. In the experiment, nearly 100 websites were selected for the flow feature extraction and recognition based on one-dimensional convolutional neural network. The recognition accuracy rate is 87.5%, indicating that this method can effectively fingerprint TOR anonymous communication.

In the process of applying deep learning to intrusion detection, in order to ensure the recognition accuracy of the model, a large number of data sets need to be classified manually, and then the model training is carried out after labeling. In practice, the efficiency of manual label designation for enough data sets is extremely low. This paper aims at the experimental data set encountered in the intrusion detection of intelligent network vehicles The problem of classification difficulty is proposed, and a method of vehicle intrusion detection based on Generative Adversarial Networks is proposed. Firstly, the vehicle driving data is collected, the collected data are put into the Generative Adversarial Networks for data classification, and the data set after training classification is used for model training. The experimental results show that the data classification can effectively improve work efficiency and reduce the resource overhead, which is practical in the application field.

Text classification is a fundamental part of natural language processing and can help with many downstream tasks, such as emotion analysis, question and answer systems, and recommendation systems. The graph convolution neural network has the natural superiority in the non - Euclidean space data. For Chinese text data, there is a lot of correlation between the data, using the graph convolutional neural network for text classification can achieve good results. In our experiment, we use a simple one-hot encoding of the word vector to process our words and use of the graph convolutional neural network can achieve 94.24% accuracy in our data set.

The current era is the era of big data and 5G. The network security data in the network is different from the past, and the network security data is growing exponentially. As an important line of defense for network security, intrusion detection technology can efficiently detect and process massive amounts of security data has become an important factor restricting its development. The feature selection method of intrusion detection data directly affects the efficiency of intrusion detection. Therefore, this paper proposes a feature selection algorithm based on pearson correlation coefficient, which performs feature specification on many features, which greatly reduces the amount of security data that needs to be processed, and effectively reduces the dimensionality of the data to increase the intrusion. Detection efficiency.

Web application brings us convenience but also has some potential security problems. SQL injection attacks topped the list of Top 10 Network Security Problems released by OWASP, and the detection technology of SQL injection attacks has been one of the hotspots of network security research. In this paper, we propose a SQL injection detection method that does not rely on background rule base by using a natural language processing model and deep learning framework on the basis of comprehensive domestic and international research. The method can improve the accuracy and reduce the false alarm rate while allowing the machine to automatically learn the language model features of SQL injection attacks, greatly reducing human intervention and providing some defense against 0day attacks that never occur.

Under the background of high-speed development of the power grid, the traditional manual detection method is low efficiency and high cost, which are not able to meet the current requirements. With the development of UAV technology, the method of using UAV to inspect transmission lines has a positive effect on improving detection efficiency and quality. In recent years, deep learning has developed rapidly, which provides a new solution for the analysis and processing of UAV aerial inspection images. This paper focuses on the deep learning algorithm to study the insulator location and fault identification in aerial photographs, which plays an important role in realizing the automation and intelligence of UAV detection.

With the development of Internet of Things technology, cloud computing technology, big data technology, and artificial intelligence technology, the application of these new technologies to rural life can help rural elderly care at home. The home care system for the elderly in rural areas is based on artificial intelligence technology and uses the Internet of things technology to collect real-time environmental parameters at home and to optimize the living environment. It uses intelligent image recognition technology to judge the behavior of the elderly and interlopers and gives effective early warning and alarm. It uses wearable devices to measure the health of the elderly in real-time. It uses the man-machine dialogue program to solve the life needs of the elderly to improve the well-being of the elderly life. It integrates big data and cloud computing technology to process the collected information.

The Internet has brought us convenience, at the same timeimportant personal information has been exposed. Once the information is leaked, it may cause huge economic losses to individuals. Therefore, it is urgent to realize network security. The traditional method of detecting phishing websites is aimed at big data. Due to the relatively large amount of data, it indirectly causes low efficiency and often fails to solve user needs in a short time. In the detection, the operation is actually performed on the small data stream. If the small data is fundamentally classified, the efficiency will be greatly improved. In summary, based on the realization that we can identify phishing websites, we mainly solve the problems of speed and accuracy. Therefore, this paper uses the R-SVM algorithm to study the data to reduce the time spent on detecting samples and improve efficiency.

With the development of new generation information technology such as artificial intelligence, Internet of things and big data, traditional agriculture has changed and gradually changed to intelligent agriculture. In intelligent agriculture, sensor technology can collect all kinds of information in agricultural production, communication technology can transmit the collected information, and use big data analysis to guide agricultural production and agricultural product sales, The application of artificial intelligence makes agricultural production high-yield and high efficiency, and unmanned driving technology can realize automatic production and fine management. This paper elaborates the application of these key technologies in agriculture.

In early 2020, the new crown pneumonia epidemic swept in, in order to block the spread of the epidemic to the campus, the Ministry of Education issued the "spring semester delayed school opening" request, "stop classes do not stop" the voice immediately rang. At one time, how to combine modern information technology with the teaching content of various subjects to maximize the efficiency of online teaching has become the primary problem for every teacher to think about, and in this environment, the dance educators who are mainly based on the traditional teaching form of "oral teaching" also ushered in a great test. By analyzing the advantages and disadvantages of traditional dance teaching methods and based on the author's online teaching practice activities for one semester, this paper makes a comprehensive analysis of the application value and significance of modern information technology in Chinese dance teaching, in order to further enrich the Chinese dance teaching mode, optimize the dance teaching resources, make the dance teaching methods and contents more in line with the requirements of China's modern technology, and provide theoretical basis and practical operation guidance for the future modern dance teaching mode.

Aimed at the improvement of the preparation accuracy of the tobacco paper reinforcing agent, the concentration control of the tobacco paper reinforcing agent was considered. The concentration of reinforcing agents is affected by many factors, such as influent flow, powder feeding speed, external interference, and so on. The compounding process has typical nonlinear, multivariable, strong coupling characteristics, and the traditional control method has a poor effect. In order to improve the quality of tobacco paper pulp, the linear auto disturbance rejection controller model of enhancer was designed to achieve the purpose of active disturbance resistance and improve the control effect of enhancer concentration. Through three groups of simulation experiments, simulation of ADRC to concentration tracking, concentration tracking under disturbance, and comparison with traditional PID control, the linear ADRC is capable of estimating and compensating well. The simulation results show that the linear ADRC has a good anti-interference performance. At the same time, the PLC control system and human-computer interaction software are designed to realize the monitoring of each process parameter in the preparation process of the reinforcing agent, which provides a practical approach for the concentration control in the preparation process of reinforcing agent.

At present, the main application of artificial intelligence in the media content production industry is to greatly reduce the application threshold, to a certain extent, to enable media workers to collect data from the low-end and produce visualization works for media terminals free from repetitive labor. Such as visual effects production, special effects design, video editing, subtitles sorting, data collection and comparison, image-text audio-video conversion, host broadcasting, and press release writing, are now increasingly completed through artificial intelligence programs. The working procedures of media content production are becoming more and more automated and intelligent.

For the problem that it is difficult to recognize the type of link establishment (LE) behaviors of radio station whose communication protocol is unknown, a method is proposed to solve the problem by using convolutional neural network (CNN) to recognize LE behaviors. Themethod processes physical layer signals directly and breaks through limitation of unknown protocol standard. Three classical CNN models are optimized through experiments so that they become more suitable for recognition of one-dimensional time series signals. Experimental results show CNN can effectively recognize the different link establishment behaviors with a large number of training samples. Moreover, DenseNet whose recognition accuracy can reach 96% when the signal-to-noise ratio (SNR) is 0dB has the best performance compared to GoogLeNet and ResNet.

cenarios, studies the construction of the digital test system based on intelligent glasses, based on the standard operating procedure operation guide, inspection data online and conformity intelligent decision, heterogeneous data packet generation and analysis of key technology, realized the application of spacecraft product inspection process, improves the working efficiency and quality.

Thai as a low-resource language has a large word segmentation performance improvement space. In this paper, we investigate a sequence-to-sequence model for Thai word segmentation with two different recurrent neural networks, which could transform one input sequence into another output sequence. Furthermore, we evaluate datasets in four different fields compared then with other multiple word segmentation models, and the F1 value in the encyclopedia dataset reaches 97.15%. The results show that the proposed model has superior performance and is more effective, it is worth mentioning that the expected results can be achieved even with limited data resources.

In agent control issues, the idea of combining reinforcement learning and planning has attracted much attention. Two methods focus on micro and macro action respectively. Their advantages would show together if there is good cooperation between them. An essential for the cooperation is to find an appropriate boundary, assigning different functions to each method. Such a boundary could be represented by parameters in a planning algorithm. In this paper, we create an optimization strategy for planning parameters, through analysis of the connection of reaction and planning; we also create a non-gradient method for accelerating the optimization. The whole algorithm can find a satisfactory setting of planning parameters, making full use of the reaction capability of specific agents.

Convolutional neural networks are widely used in human production and life, but due to their large amount of calculation and complex calculation mode, their calculation speed is slow, so it is necessary to design a dedicated hardware accelerator. This paper firstly analyzes the algorithm of the convolutional neural network and decomposes the algorithm into multiple basic operations. For the convolution operation with the largest amount of calculation and complex operation mode, a near calculation storage array is designed according to its operational characteristics. Furthermore, a convolutional neural network accelerator architecture is proposed to realize the fast operation of a convolutional neural network.

Based on the network security protection of the power monitoring system, this paper proposes a pattern recognition model for the operation behavior of the operation and maintenance personnel of power monitoring system. Through the technical research of the alarm pattern recognition of the network security management platform, and carries out the operation type data of the network security management platform Clustering analysis uses historical data to intelligently train the clustering model. Through training, the risk level classification of the operation behavior is obtained, and then real-time data is introduced into the model for detection, which can judge the degree of risk of the operation event in real time. With the improvement of the K-means clustering algorithm and the application of big data analysis, it further improves the he intelligence level of the platform and provides technical support for the intelligent protection of network.

In this paper, we propose a new improved algorithm based on the deep integration of GoogleNet and Residual neural network and we call it GRSN. The new improved algorithm has the new advantages of multi-size and small convolution kernel in the same layer in the network and the advantage of interlayer hop connection (bypass) to reduce information loss. The algorithm is applied to the general image data set cifair10, and the experimental results are compared with that of GoogleNet, the accuracy is improved, the convergence is accelerated, and the stability is better.

Aiming at the low efficiency of vehicle object detection in real scenes, this paper proposes an improved RetinaNet. An octave convolution structure and a weight pyramid structure are introduced respectively to improve the detection performance of RetinaNet for vehicles. Specifically, we use octave convolution instead of the traditional convolution layer to improve the feature map's representation of detailed information. In addition, in order to improve the quality of feature fusion, a weighted feature pyramid network (WFPN) structure is proposed to limit the propagation of gradients between different levels. The experimental results on the DETRAC data set show that the method has good detection results for vehicle targets of different scales in different scenarios, and can better meet the needs of practical applications.

In the age of the Internet, while the network is sending a lot of information to people, hackers also transmit a lot of malicious code through the network. Hackers use these malicious codes to steal sensitive information from infected people and damage machines and devices to achieve their evil goals. Therefore, it is very important to accurately detect malware to protect users from loss. There are now three major methods of detecting malware, which are signature-based, behavior-based, and heuristic-based methods. However, with the rapid increase in the types and number of malware, signature -based method can't detect unknown malware and behavior-based cannot guarantee the False Positive Ratio. So these detection methods can no longer meet the needs. Therefore, some researchers proposed some heuristic-based detection methods. In this article we overview the methods used to extract features and the features extracted in heuristic detection and discuss the advantages and disadvantages of the features

Accurate prediction of product quality can be used to control the online production process of workpieces, an appropriate prediction model can effectively improve the yield of industrial production. In order to obtain missing data, the classic method is to regress them one by one. However, the traditional regression prediction method destroys the correlation of the data and cannot use the quality inspection index efficiently. This paper presents an improved hybrid model. The essence lies in a novel easily pluggable CGAN module that refines the previous step of regression data. compared with the regression method, the CGAN module can learn the joint d istribution of data. The experiment of real data result shows that our method has high accuracy compared with the existing ones

The online comment for doctors plays an important role in assisting patients to master the real medical situation and choose medical treatment, which greatly reduces the adverse effects of online medical market information asymmetry. but because ordinary patients lack professional medical knowledge and can not accurately and efficiently write comments that are useful to similar patients, we identify useful topics from online medical reviews and make combination recommendations based on patients with different diseases. guide patients to write comments from the most useful dimensions to alleviate the problem of information overload, thus maximizing the limited human resource utility. This study aimed at online medical platforms such as good doctors and micro-doctors collected online doctor reviews for major diseases and used LDA topic mining and fsQCA fuzzy set qualitative comparative analysis to analyze key topics that affect the usefulness of reviews and optimal topic combinations under different disease types.

In this study, a novel signal modulation recognition framework has been proposed for automatically classifying eleven different modulation types with various SNR values. The framework employs both the raw complex-valued I/Q signal and its time-frequency description to represent the radio signal. And, a hybrid deep neural network is presented to recognize different modulation types from the representation data by leveraging the appealing properties of a convolutional neural network (CNN) and a long short-term memory (LSTM) network. Extensive validation of our scheme is performed on a large public dataset by comparing it with three existing 。 methods from literature, and our scheme yields quite promising results in terms of recognition accuracy.

Deep reinforcement learning (deep RL) achieved big successes with the advantage of deep learning techniques, while it also introduces the disadvantage of the model interpretability. Bad interpretability is a great obstacle for deep RL to be applied in real situations or human-machine interaction situations. Borrowed from the deep learning field, the techniques of saliency maps recently become popular to improve the interpretability of deep RL. However, the saliency maps still cannot provide specific and clear enough model interpretations for the behavior of deep RL agents. In this paper, we propose to use hierarchical conceptual embedding techniques to introduce prior-knowledge in the deep neural network (DNN) based models of deep RL agents and then generate the saliency maps for all the embedded factors. As a result, we can track and discover the important factors that influence the decisions of deep RL agents.

With the rapid development of the 5^thgeneration of wireless network communication technology, a high-speed data rate, broader bandwidth, and low-latency wireless network has already come to the truth. Meanwhile, Edge Computing which put data processing at the edge of the network has already been well developed, and it has changed the computing mode tremendously. And also, Artificial Intelligence (AI) has made a great breakthrough from Computer Vision, Unmanned Vehicle to Natural Language Processing. Nevertheless, with the increasing model size and model depth, the training data set and communication delay has become the bottleneck of AI popularization. As zillions of bytes of data are being generated at the network edge, and also lots of AI applications are being utilized by enormous customers lied at the edge, the combination of Edge Computing and AI is imperative. This paper has done a deep investigation into the new field about the confluence of Edge Computing and AI, aiming of discussing the concept, architecture, and research ideas on Edge Intelligence. To the end, this paper provides the road-map for future researching work.

Slot Filling (SF) is a critical part of spoken language understanding (SLU) which targets to capture semantic constituents from a specific utterance. It is considered as a sequence labeling issue. Currently, recurrent neural networks have shown promising effectiveness in this issue. Considering the effects of interrelated information within adjacent words and labels, we present that a novel approach consists of bi-directional gate recurrent unit (Bi-GRU), attention mechanism, and conditional random field (CRF). Our model can utilize interrelated information from words in the neighborhood, highlight key information, and exploit the dependencies within labels corresponding to surrounding words. The empirical experiments illustrate that our model significantly boosts the F1 score with around 1% and 5.1% relative enhancement on two public benchmark dataset ATIS and SNIPS separately.

A deep reinforcement learning algorithm has been widely used for learning dialog policy in a task-oriented dialog system. Dialog agent collects training data and improves its policy by interacting with users. Interact with real users is time-consuming and not realistic. Consequently, we usually build a user simulator instead of a real user. At the beginning of each dialogue, the simulator will sample a user goal extract from training data. Then the simulator will communicate with the dialog agent to accomplish this user goal. The existing user simulator usually samples this user goal randomly. This will cause the dialog agent to waste a lot of time to learn what it already learned. To solve this problem, we propose two user goal weighting methods which give relatively large weight to the user goal the current dialog agent can't accomplish. This lets the dialog agent pay more attention to those user goals. Experiment result in a movie-ticket booking task shows that the proposed weighted user goal sampling method can effectively accelerate the policy learning progress compares to the random sampling method.

The work aims to solve the problems of low detection accuracy, and poor reliability of magnetic material surface defects. A lightweight multi-scale feature ResNet (LMSF-ResNet) method was proposed in this paper. The method reduces the calculation and parameters of the neural network by using group convolutions and channel split operator and merging features from different branches of distinct scales by using the multi-scale model's structure. The algorithm employs the strategy of using the channel attention model to improve accuracy at the same time. This paper takes the defect images of rare earth magnetic materials as an example to verify its effectiveness. The results show that the method has higher detection accuracy and stronger stability, and can be widely used to accurately detect surface defects of magnetic materials.

As a typical representative of information filtering technology in the era of big data, a recommendation system is an important means to solve the problem of information overload. A collaborative filtering recommendation algorithm is one of the important technologies to realize the recommendation system, but the traditional collaborative filtering algorithm only considers the similarity of ratings between users. As the number of users and the number of items increases, it faces user interest drift and reduced recommendation Precision, and other issues. In this regard, a collaborative filtering algorithm based on improved time function and user similarity is proposed. First, considering that user interests will dynamically change over time, this paper introduces an improved time function in the traditional scoring similarity; secondly, considering the impact of the number of items evaluated by users on the calculation of similarity measurement, this paper takes the Pearson correlation coefficient weighted scoring into account; Finally, the fusion recommendation is based on the improved time function and the weighted Pearson correlation coefficient to improve the Precision of recommendation prediction. This paper conducts simulation experiments on MovieLens 100K and 1M data sets respectively. The results show that, compared with the traditional collaborative filtering recommendation algorithm, combining the improved time function and the weighted Pearson correlation coefficient can effectively improve the recommendation Precision.

Ground penetrating radar (GPR) is widely used in the field of intelligent road detection because of its non-destructive detection method, which is based on an electromagnetic wave reflection mechanism. However, this method requires large-scale data processing and also relies on manual judgment, which is time-consuming and laborious. To solve this problem, this paper analyzes and studies the GPR images of underground pipelines and uneven settlement of urban roads by means of road surface measurement and laboratory tests. The radar image data set is constructed by collecting radar images and denoising and marking them. Then, Deep Feature Selection Net is adopted to improve the fast region-based convolutional neural network (Faster R-CNN) to enhance the network's ability to extract features from radar images. Finally, by comparing with the improved faster R-CNN model, it is found that the automatic identification rate of underground pipelines and uneven settlement in urban roads increases, reaching more than 80%.

In this paper, we propose an integrated method for lane detection and vehicle detection, which tries to make a real-time analysis of vehicle video for identifying lane, detecting, and tracking forward vehicles. In lane detection, image preprocessing, line detection based on improved Hough transform, and straight-line model reconstruction are used. For vehicle detection, preprocessing, vehicle shadow merging based on the improved search algorithm, regions of interest (ROI) demarcation, lane determination, and vehicle tracking are used. The experiment results show that the time it takes to process an image is about 25ms. Additionally, the lane detection rate of vehicles driving on a structured road is approximately 98%, and the vehicle detection rate of the closest forward vehicle is approximately 81%.

A data-driven flood forecasting model based on a convolution neural network is proposed for the small and medium-sized watershed. Pearson correlation coefficient analysis was used to determine the time length of the input data. The historical rainfall and discharge were used to create the two-dimensional input data matrix. NAS was used to determine the structure of the model. The experiment results in Tunxi watershed in Anhui Province show that the accuracy is high under the 1˜3h forecast period, with the forecast period increased, accuracy would decrease.

It is challenging to find an effective license plate detection and recognition method due to the different conditions during the image acquisition phase. This paper aims to develop a new accurate and efficient method based on color difference and SVM recognition model that yields better performance for vehicle images under low quality. The proposed method is tested with 200 images which involve many difficult conditions, such as low resolution, night-lighting, dirt, complicated background, and distortion problems. The experimental results demonstrate very satisfactory performance for license plate detection in terms of speed and accuracy and are better than the existing methods like edge detection or HSV color conversion method. The overall probability of localization is close to 100%, with a false recognition rate of 2%.

Recent years have witnessed the boom of artificial intelligence. With the advancement of judicial information, using artificial intelligence (AI) technologies to mine judicial big data is of great significance in smart courts. However, the reasoning of the evidence chain mainly relies on the judge's manual work in the litigation process. How to model the multi-source evidence association (MSEA) and reason credible evidence chains (CEC) automatically is largely unexplored. In this paper, we propose an MSEA model based on the Bayesian network. Firstly, we construct an MSEA network in which each evidence element serves as a node, and the node correlation probability is calculated via the association relationship among the evidence elements. Subsequently, with the guidance of the event judgment chain, the MSEA model is constructed based on Bayesian networks. In the end, we use a genetic algorithm to optimize the Bayesian network and select credible evidence chains. To the best of our knowledge, this is the first time that using a probability graph model to mining the association of multi-source evidence. Experiments and the case analysis prove the effectiveness of our method.

Facial expression recognition is widely used in various research fields. For facial expression recognition problems, deep neural network methods have a complex structure and poor interpretability, while traditional machine learning methods have less plentiful diverse features and low recognition rates. Therefore, a new Multilayer Convolution Sparse Coding (MCSC) method is proposed for facial expression recognition. The MCSC method deeply extracts the salient features of the human face through a convolutional neural network. Furthermore, it uses a multilayer sparse coding to learn layer by layer to recognize different facial expression features based on sparse coding, which improves the recognition accuracy of facial expressions. Finally, the MCSC method was validated on three public facial expression datasets, i.e. JAFFE, CK +, and Fer2013. We also compared and analyzed 5 feature extraction approaches. The results show that MCSC has the best facial expression recognition performance in the comparison algorithm. Its accuracies of the three data sets reach to 90.8%, 98.2%, and 72.4%, respectively.

Semantic matching is a hot topic in the field of NLP. It is widely used in information matching, machine translation, intelligent Q&A, and answering system. With the rapid development of artificial intelligence, people pay more attention to friendly human-computer interaction and how to obtain accurate and effective information. Aiming at the problem of the diversity and structure of Chinese, this paper introduces the cross attention mechanism. Texts provide attention to each other, then to information retrieval. Experiments on aetc-nlp dataset show that the method is effective.

With the booming of the Internet, Web public opinion plays an increasingly important role in the stability of the network community. Therefore, the sensitive information hidden on the Internet is likely to lead to unpredictable social impact. This paper focuses on the detection of Chinese sensitive information. First, we build a corpus to train the detection model. Secondly, we apply the Bert method to the detection problem. Then, many popular NLP methods are applied to this problem to show the progress of Bert in a sensitive information detection task. Finally, we got a sensitive information detection model based on BERT with a high F1 score of 97.31.

In the era of the rapid development of computers and the Internet, e-commerce has become a part of the economy of many countries. Therefore, how to use historical data of user consumption behavior to predict user shopping intentions accurately and subsequent personalized recommendations has turned into research hotspots in the field of e-commerce. This paper conducts a basic analysis of JD's e-commerce data based on machine learning. Specifically, this article constructs user-product matrix and product and user clustering by means of text processing and clustering, as well as implements a logistic regression classifier to predict the user's purchase intention of products in a certain target category in the next 5 days. Based on the JD competition data set, this article has a prediction accuracy rate of 98%. This can help e-commerce companies make better decisions.

The traditional water resources carrying capacity evaluation system obtained by the fuzzy comprehensive evaluation method has strong subjectivity in factor selection and threshold calculation, and the factor analysis is relatively one-sided. Therefore, water resources carrying capacity evaluation system based on principal component analysis is established and its threshold is studied. First, the main influencing factors of water resources carrying capacity, including water resources endowment, society, ecology, and other resources are analyzed. Then, the water resources carrying capacity evaluation system is established by selecting the relevant indicators in the evaluation system, comprehensively considering the selection principles, and selecting the index with the highest score as the main component. Finally, different methods are used to calculate the threshold value of the evaluation index. So far, the establishment of the evaluation system of water resources carrying capacity based on principal component analysis and the calculation of the threshold value is completed. The results of the case analysis show that the established water resources carrying capacity evaluation system are more scientific and effective than the traditional system.

In a non-cooperative frequency hopping communication system, the frequency hopping network station sorting of the received hybrid signals plays an important role and becomes an active research area in recent years. In order to solve the problem that the currently widely used clustering algorithm can not achieve satisfactory accuracy. In this paper, we propose a signal sorting method for hybrid frequency hopping network stations by applying the neural network to classify the frequency hopping description words of signals. Additionally, the conjugate gradient algorithm is utilized in the neural network training process to improve the convergence speed. Simulation results demonstrate that when compared with the clustering algorithm, the proposed algorithm converges with fewer iterations and delivers better sorting accuracy, especially in a low signal to noise ratio environment.

The TextCNN model is widely used in text classification tasks. It has become a comparative advantage model due to its small number of parameters, low calculation, and fast training speed. However, training a convolutional neural network requires a large amount of sample data. In many cases, there are not enough data sets as training samples. Therefore, this paper proposes a Chinese short text classification model based on TextCNN, which uses back translation to achieve data augment and compensates for the lack of training data. The experimental data shows that our proposed model has achieved good results.

In recent years, the use of Time-frequency analysis to generate Time-frequency diagrams of vibration signals, and then the use of deep learning methods to classify them has become one of the mainstream methods for bearing fault diagnosis. However, a single Time-frequency analysis method usually cannot extract a complete vibration signal, so the accuracy of the model will be affected to a certain extent. Therefore, this paper proposes a multi-view bearing fault diagnosis method based on deep learning. The same segment of vibration signal is generated by different Time-frequency analysis methods, and then the pre-trained network is used to train the model. The Flatten layer was replaced by the Global Max Pooling (GMP) layer before the layer. The experimental results show that compared with the traditional feature fusion method, the method in this paper can not only achieve better accuracy, but also has stronger generalization.

Fire detection can effectively prevent the occurrence of fire. For the current fire detection methods, traditional image processing techniques such as grayscale image processing and feature extraction processing have poor anti-interference ability, weak generalization ability, and the detection results are more sensitive to data fluctuations. At present, based on the concept of deep learning, the proposed convolutional neural network processing of extracted image features has been widely used. On this basis, an improved YOLOv3 fire detection algorithm is proposed in this paper: The K-Means++ algorithm is used for clustering analysis to obtain the corresponding anchor boxes dimension, which reduces the error detection rate caused by the bounding boxes not matching the label; secondly, the resolution of the feature image is improved and the receptive field is enlarged; the image data are sharpened and the contrast is enhanced to make the data features more prominent. The experimental results show that the detection precision of the method is 97.7%, the recall rate is 98.5%, and the fps is 19, effectively solving the problem of high error detection rate and high missing rate of traditional image processing and general CNN networks for suspected flame objects.

Wind power generation is affected by weather and historical wind power, which presents the characteristics of instability and high volatility. Most wind power prediction models ignore physics information. In this paper, a novel combined predicting model that simultaneously considers physics information and historical information is presented to address the drawbacks of existing models. First, the physical characteristics of wind speed, wind direction, and temperature are obtained by Deep Neural Network(DNN), and time-series characteristics from historical wind power are extracted by Long Short-Term Memory(LSTM). Then, the physical features and the time-series features are fully connected for feature fusion to obtain the final time-series physical features. Finally, the short-term wind power prediction is performed according to the obtained merged features. Experimental results demonstrate that the DNN-LSTM model proposed in this paper achieves high accuracy and stability, and provides technical support for wind power system dispatch.

In the computer vision, YOLO has an important position in the field of object detection, but due to its speed limitation, it is not suitable for scenes that require extremely strict real-time performance, such as smart cameras or some mobile devices. However, inverted bottleneck layers, which are built upon depth-wise separable convolution, have been the predominant building blocks in state-of-the art object detection models on mobile devices. In this work, we propose a new lightweight algorithm Smart-YOLO based on the YOLO framework which uses inverted bottleneck blocks and deep-wise separable convolution. We also put forward a new loss function to make up for the loss of accuracy caused by the replacement of the backbone network. The results show that compared with YOLOv3, the accuracy of our model is reduced by about 21%, but achieved up to 4.5× speedup, and the model size is only about 1/8 of the original. This shows that our network is smaller, faster, and more suitable for scenarios that require higher speed and efficient

Borrowing the power of deep neural networks, deep reinforcement learning achieved big success in games, and it becomes a popular method to solve the sequential decision-making problems. However, the success is still restricted to single agent training environment. Multi-agent reinforcement learning still is a challenge problem. Although some multi-agent deep reinforcement learning methods have been proposed, they can only perform well when the number of agents is very limited. In this paper, by analyzing the dynamic changing observation space and action space of multi-agent environment, we propose a novel multi-agent deep RL method that compress the joint observation space and action space as the time goes on. The proposed method is potential for a large number of agents cooperative or competitive tasks

Aiming at the problem of the intercept game in the inertial space, this paper builds a model of basic actions of the two sides of the intercept game in the inertial space, explored the applicability of Deep Reinforcement Learning in inertial space game problem-solving. Based on Proximal Policy Optimization (PPO), this inertial space game problem is solved and the optimal solution is obtained by the reward designing with cumulative miss distance and minimum distance respectively, and finally, the effectiveness of the algorithm based on PPO is verified through the game simulation and results from comparison

As a common biometric recognition technology, face recognition is also an important research direction in the field of computer. Although compared with the initial research, the current research has made great progress, but there are still many difficulties in practical application. In this paper, by extracting HOG features, after introducing the detailed steps of PCA and LDA subspace feature extraction methods, dimensionality reduction feature extraction method combing PCA with LDA is applied to extract face features. This method first uses PCA to reduce the dimension of face features, and then uses LDA for linear discriminant analysis. Finally, the feature extraction methods based on PCA and LDA are tested and compared in FERET standard face database and CAS-PEAL database of Chinese Academy of Sciences

At present, facial expression recognition technology is widely used in artificial intelligence, transportation, medical and other aspects, so it has important research value. Traditional facial expression recognition uses manual feature extraction method with low accuracy and weak generalization ability, which is difficult to be applied in real life. With the development of deep learning, convolution neural network appears in people's vision. Different from traditional manual feature extraction, convolution neural network can learn image features independently, and learn more features. In addition, it has the advantage of sharing the weight with the neural network. Although convolution neural network has multiple advantages, it also has some disadvantages, especially over fitting. In this paper, the model of convolution network is improved based on the classical VGGNet according to the working principle of convolution neural network. In this new model, the number of convolution kernels is reduced in parameters, and the global average pool layer is used to replace the full connection layer in the structure, so as to reduce the degree of over fitting and decrease the operation parameters. Finally, experiments show that the accuracy, generalization and consumption of resource are enhanced in the new model. It is proposed that the new method is better than the traditional convolution network VGGNet.

The prevention of d isease, has the important guiding significance to safeguard the health of the people. In order to improve the accuracy rate of mumps disease forecast of Yinchuan City, Ning xia, joined the combined kernel function idea based on regression prediction model of the traditional Support Vector Machine, the polynomial and radial basis kernel function combination instead of the traditional single kernel function model, proposed a combination of kernel function of support vector machine algorithm, with mumps Ning xia disease for the practice of Yinchuan City, to conduct research and forecast for the pathogenesis of rules combined with the local weather changes. The meteorological data and the number of diseases using correlation analysis method for processing, select the factor related coefficient ordering large from, then these factors should be selected using a single kernel and mixed kernel function for processing, training and prediction using support vector machine. The experimental results show that the mixed kernel function of Support vector machine prediction model results than the single kernel function regression is better

In order to alleviate the multi-scale problem caused by the scale change between object instances, pyramids are widely used in target detection. Although these target detectors with characteristic pyramid structure have achieved good results, they have some limitations because they simply construct characteristic pyramids according to the skeleton multi-scale pyramid structure originally used for target classification tasks. In this study, based on M2Det, a multi-scale feature with richer multilevel information is proposed to construct a more effective feature pyramid to detect targets of different scales.First, the basic features with multi-level features will be fused and extracted from the backbone.Then, the basic features are input into the M module, and the features generated by each P module are used as the features of the detection object in the form of dense connection. Finally, the attention mechanism is introduced into the L module to assemble the features with the same scale and construct a feature pyramid for target detection. On the COCO dataset, MLPNet implements the results of AP37.8.

According to the characteristics of CP-OFDM signals, analyzed and studied the subcarrier number estimation, timing synchronization and frequency offset compensation algorithms of CP-OFDM signals under non-cooperative reception conditions, and verified the feasibility of these algorithms based on UAV image transmission signals. As the results show, the algorithm proposed in the article can complete blind demodulation of the CP-OFDM signal and finally complete the sensing and recognition of small UAV signals under the condition of unknown CP-OFDM signal format by means of analysis and estimation algorithms.

Traditional recommendation methods decompose keywords and key sentences less frequently, resulting in low accuracy of recommended content. Therefore, a personalized recommendation method for English teaching resources based on artificial intelligence technology is proposed. Web mining technology is used to collect user personalized data and mine teaching resource rules and patterns that can represent user characteristics. Artificial intelligence technology is used to simulate the intelligent behavior of users searching for resources, to obtain words with a higher frequency. The words with higher relative contribution are selected as keywords, thereby calculating the similarity between keywords and user characteristics, then the teaching resources with high similarity are taken as the recommendation target, and the recommended format is selected to implement resource recommendation. Experimental results show that compared with traditional methods, this method improves the accuracy of recommended content and enables teaching resources to better meet the individual needs of users.

For the deep strata with fracture zone, fault zone and various permeability and fracture loss, the geological conditions are particularly complicated, which makes it particularly difficult to predict the safety density window and engineering risk, through investigation and research, combined with the actual situation of a deep oil and gas reservoirs in northwest China, for the formation of three pressure section in the regions was determined, and through the actual leakage occurs correction of formation fracture pressure of drilling fluid density as a standard of lost circulation project risk evaluation, the author combined with big data, based on a certain algorithm, got the regionalization of the region narrow safe density window, on narrow density window combination of generalized stress and strength interference theory, based on the real-time drilling fluid density in the region one of the key Wells project risk evaluation, the project risk is match with the actual project risk situation

Today's electromagnetic environment is still not optimistic enough, and spectrum resources are relatively few. In the actual allocation, unevenness will inevitably occur. In addition, the current level of intelligent monitoring is limited, and it is impossible to fully grasp the dynamics of frequency use. Spectrum monitoring emergency maneuverability still needs to be further improved, and there is a lack of refined spectrum resource management measures. In order to solve this problem as soon as possible, the article proposed a solution to the electromagnetic spectrum monitoring problem based on Cloud Computing, Big Data and Artificial Intelligence technology, proposed construction plans for various main applications and corresponding monitoring services, mainly strengthening the construction of handheld monitoring systems, electromagnetic spectrum monitoring and control systems, cloud monitoring systems, Big Data analysis systems, and intelligent monitoring and dispatching systems.

Agricultural data is typically characterized by at least 3 characteristics: massive data volume, various data sources, and complex structure. With the popularization of next-generation information technology, precision agriculture has become an emergent development direction in agricultural informatization. This paper proposed a big data application framework for precision agriculture. The model includes data sources, data integration, and data analytics. Use of big data application provides better guidance for precision agricultural.

Many functional areas have been formed in the development of the city, and the gathering of taxis can basically reflect the travel situation of the crowd. Combining urban functional areas and taxi GPS data to analyze crowd travel conditions can provide decision support for the construction of smart cities. Therefore, this paper analyzed the travel situation of crowd in Changsha urban area based on multi-source data. First, this paper clusters the distribution of public transit stations in Changsha, and then imports the geographic location information of the cluster centers into the Voronoi diagram in ArcGis to divide the urban area of Changsha. Then the divided areas and Point of interest (POI) data are counted and analyzed to obtain the functionality of each area. Finally, the taxi GPS data is clustered according to the time period, and the cluster center is mapped to the urban functional area, to analyzed the crowd clustering situation of each functional area in Changsha in different time periods.

Aiming at the problems of big data resources of investment statistics in the management process, such as unstructured, massive, and multi-type, the research on the big data integration method for investment statistics based on artificial intelligence technology is conducted. A new method of big data integration is proposed through the correlation analysis of investment statistics big data, the retrieval of investment statistics information resources, and the autonomous integration of big data based on artificial intelligence technology. Experiments have proved that compared with traditional integration methods, this method can effectively reduce the loss of information resources, ensure the integrity of the integrated investment statistics big data resources, and provide theoretical and technical explorations for the compilation of enterprise investment statistics, also provide a reference for big data integration in other related fields.

Exploring unknown miRNA-disease associations by computational tools is a new way to study the correlations between genes and diseases. in this paper, we proposed a new model named ABPUSVM, which consisted of a new sample strategy based on Positive-Unlabeled learning and a prediction model that combined AdaBoost and SVM. When ABPUSVM was applied to predict unknown associations of miRNA-disease, the AUC of 0.9383 improved greatly based on 5 folds cross-validation and showed its excellent performance, which indicated ABPUSVM was significantly better than other classic models. Afterward, a case study confirmed that 46 out of the top predicted 50 miRNAs of breast cancer by ABPUSVM were supported by three databases. whose results showed the dataset by ABPUSVM were significantly better than that of the other methods. All results have shown that ABPUSVM is a promising and potential tool for exploring the associations of miRNA-disease.

The continuous variable quantum key distribution protocol (CV-QKD) has the advantages of easy generation of quantum signals, convenient measurement and high communication capacity. However, the multi-dimensional data reconciliation is mainly implemented by binary LDPC code as quantum error correction code, which is greatly affected by complex environments and the bit error rate is high. In order to solve this problem, acontinuous variable quantum key distribution protocol based onmulti-dimension data reconciliation with Polar code, that can be applied to long-distance transmission in power systems, is proposed. First, we prepare an EPR quantum state, and perform heterodyne detection on one of the quantum states, and obtain the values of two orthogonal components at the same time which is the coherent state of the information carrier. Then the continuous variable quantum key distribution is realized by six steps (i.e., quantum state preparation, gaussian random detection, data filtering, calculation of bit error rate, data reconciliation, privacy amplification).In data reconciliation, we use Polar code as the error correction code and use reverse coding technology to achieve key error correction, which can effectively improve the negotiation efficiency. Numerical simulation verifies the feasibility of the protocol, and the protocol is more efficient than the multi-dimensiondata reconciliation protocol based on LDPC code under the same conditions.

This study concerns a simulation and optimisation of the personnel evacuation modelling of a shopping mall in Weiyang District, Xi'an. First, investigations were conducted to determine the human flow rate tand occupant distribution during the rush hours to builda queuing theory model. Then, a simulation was implemented with an evacuation simulation software Pathfinder, followed by an analysis to the evacuation status and bottlenecks. On the basis of the evacuation status, proposals were put forward on how to overcome the bottlenecks to achieve the optimal matching among the personnel, emergency staircases and aisles. Finally, the improved scheme was simulated. Compared with time taken by the evavuation before improvement, the time taken was greatly shortened after improvement, proving the feasibility of the scheme. Through the simulation experimentaiton, it was found that ecacuation modelling can greatly facilitate reseaches on the evacuation problems in multistoryed shopping malls.

Agriculture and forestry are the source of food, clothing and subsistence for human beings and also the primary conditions for all production. The forecast of the gross output value of agriculture and forestry directly affects the purchasing, pricing and storage of important plans in the agricultural sector. This paper collects quarterly data from 2012 to 2020 in Guangxi Zhuang Autonomous Region and uses the Holt-Winters prediction model based on the multiplication to fit the historical data. The model was validated by calculating the optimal smoothing coefficient through the minimum residual sum of squares, making predictions and comparing actual values for the data from 2018 to 2020. Finally, the model was used to derive future forecasts for agriculture and forestry from 2020 to 2022, and recommendations for agricultural development were made based on the forecasts.

Aiming at the difficulty of load forecasting due to the current holiday load jump, a holiday load forecasting method with multi-scale feature combination is proposed. Feature extraction and recoding of date information, load information and weather information, and effectively use of historical data information, reconstruct load forecast feature combination. This method is used to reconstruct the electric load data of a certain area in Jiangsu Province. XGBoost and LSTM were used to predict the holiday load in the reconstructed multi-scale feature combination dataset and the traditional feature combination dataset. The experimental results show that in both load forecasting models, this feature combination method can effectively mine the latent relationship contained in historical data, represent more refined prior knowledge, and improve the accuracy of holiday load forecasting.

It is difficult for the traditional topic relevance algorithm based on word frequency and probability statistics model to deal with the ambiguous topic of search keywords, which leads to the retrieval results containing much information that users are not interested in. To solve this problem, a page-topic relevance algorithm based on BM25 and paragraph-semantic correlation is proposed in this paper. The semantic correlation of one page is calculated by the paragraph-semantic classification using a pre-trained deep neural network model, and is weighted with the BM25 retrieval score. Experimental results show that compared with the traditional BM25 algorithm, this algorithm can effectively improve the retrieval accuracy.

This article, the official news site of Linyi five services in Shanghai, Yiwu GanZhou Shenzhen national logistics hub of news as the data source, through Word2Vec and construction LSTM emotion classification model, with positive or negative emotion in general categories, calculating to analyze its emotional value, from the emotional category and time series analysis and word frequency vector to Linyi public opinion analysis of logistics hub

News is an important form to reflect current politics, which attracts people's attention. The emergence of crawlers provides a convenient way for people to obtain useful information from mass news. Through Python language, this paper takes the relevant news of business logistics hub as an example to carry out the experiment of network news crawling. This paper introduces the workflow, design and implementation of the crawler in detail. The experiment proves that the designed crawler can get news quickly and can provide people with the information they need.

Microblog comments express the public's views and attitudes on Microblog hot topics and hot events. Therefore, by excavating the theme of Microblog comments, we can help people understand the trend of public opinion and the attitude of the public towards the event. In view of Microblog comment topic mining, the problem we need to solve is how to use simple methods to quickly and efficiently obtain Microblog comment content, and how to use what methods to mine the theme of Microblog hot events to a great extent. First of all, we construct dynamic links through static web pages, write Microblog comment web crawler program in Python language, and solve the problem of crawling Microblog dynamically loading web page comments. Secondly, we use the Latent Dirichlet Allocation model, which is the implicit Dirichlet distribution, to mine the text.

When shopping on online e-commerce platforms as they can only contact products through photos and other means, it is easy to have cognitive deviations on the products on online platforms. This paper makes intelligent acquisition and analysis of commodity review data by using web crawler tools. Through a series of operations such as comment data capture, data preprocessing, Chinese word segmentation, emotional tendency analysis, LDA model analysis, it analyzes buyers' emotional tendency of this commodity to help consumers choose the commodities they really like.

At present, there are many methods for data dimensionality reduction, but most of the results of decomposition methods allow negative values. Obviously, these negative values have no physical meaning in practical problems. The non-negative matrix factorization method is a dimensionality reduction under the condition of ensuring non-negative values. This paper mainly uses a non-negative matrix factorization algorithm to perform dimensionality reduction representation and local feature extraction of data, and then classify. Experiments show that the algorithm in this paper is reasonable and effective.

Web crawlers are an important part of modern search engines. With the development of the times, data has shown explosive growth, and mankind has entered a "big data era". For example, Wikipedia, which carries knowledge achievements from all over the world, records real-time news that occurs every day and provides users with a good text search database[1]. Wikipedia updates data up to 50+GB every day. This project focuses on solving the problems of data acquisition and data analysis. At the same time, it downloads and parses the latest data of Wikipedia and analyzes XML files, and then uses SVM algorithm and Naive Bayes algorithm to classify articles, Train the model to download Wikipedia files efficiently and parse XML files.

In recent years, researchers at home and abroad have accumulated a lot of experience in the research of data visualization technology, and they have played animportant role in scientific discovery, medical diagnosis, business decision-making, and engineering applications. As a library developed using Python language, Matplotlib has a concise language, high drawing accuracy, and simple and easy-to-understand code. This article first introduces data visualization and related technologies used and then uses Python's Matplotlib library and pyecharts library to realize data visualization. Through representative examples, combined with the use of correct charts, visual processing of data in different fields, so as to further analyze the effect of visualization.

One of the important reasons for irrational drug use is that the medical industry has been facing the problem of complex and redundant information. In order to solve this problem, a multitude of researchers carried out research on rational drug use by data mining technology and tried to apply it to real medical treatment. Apriori is one of the most frequently used and valuable algorithms. Based on the Apriori algorithm, we discussed the basic theory and the improved result. We elaborated on the detail and analyzed the development of the applied scenarios in modern medicine and traditional Chinese medicine. Then we discovered the applied scenarios of the Apriori algorithm mainly included medication law, dose research, and experience inheritance. Through the analysis of the literature on the use of the Apriori algorithm in rational drug use and the current development trend, this paper pointed out the existing problems in this field and put forward the future work and research focus.

With the rapid development of information technology, big data is gradually being popularized and promoted in various industries. Smart wearable devices have attracted much attention for their objective, accurate and continuous measurement of the amount of exercise. With the continuous improvement of intelligent equipment performance, large data analysis and intelligent wearable device technology can real-time capture the movement of the students in class data: trajectory, heart rate, speed, cost, etc., through data analysis, to monitor and intervene student's sports activities in the class at the same time, can help the teacher more scientific system of the training plan, a more comprehensive understanding of the students in the class level and performance, enhance the level of student's physical stamina and physical health and will bring far-reaching influence to sports research.

In recent years, the sources of maritime spatial-temporal trajectory data have been increasing. With the widespread application of detection equipment such as radars, ship-based platforms, and satellite, maritime trajectory data has shown feature of volume, variety, velocity and valueless. The increase in data types and amount of data has led to complexity and difficulty of data processing. Traditional data mining algorithms are facing serious difficulty. This paper reviews researches for maritime spatial-temporal trajectory mining in recent years on trajectory clustering, anomaly detection, and trajectory prediction. Difficulties current researches faced are pointed out and prospects of deep learning are discussed in this paper.

IOT devices are easy to be attacked by hackers because of their intelligence and openness. The vulnerability mining of IOT devices has become a key research field in the security field. This paper studies a method of vulnerability mining based on function hook mechanism, which can make AFL QEMU mode is suitable for the binary fuzzy test of firmware CGI program of IOT devices, it can effectively test the firmware CGI program. Through the experiment, we successfully found the common vulnerability of IOT router equipment and through the reverse analysis of firmware, we trace back to the assembly code of the corresponding vulnerability point, and then carry out the corresponding analysis.

In order to solve the problem that it is necessary to scan the database many times and produce unnecessary frequent itemsets in the process of mining indirect association rules, a new algorithm FPI-mine based on FP-Tree is designed to mine the indirect association rules in transaction database. Firstly, FP-Tree is constructed, and then the indirect item pairs and intermediate support sets of all frequent items are found. Finally, all the indirect association rules are obtained by mining algorithm. It can directly mine indirect association rules without generating all frequent itemsets. Finally, the effectiveness of the algorithm is verified by experiments.

In this study, movie reviews are used as data sets to extract related phrases, topics, and sentiment scores from the text. Based on users' information, users' behavior preferences and their influences are analysed, and text semantic information is mined from multiple perspectives. A variety of data processing and machine learning methods including text segmentation, Apriori association rule mining algorithm, sentiment analysis, linear fitting, TFIDF algorithm, PCA dimensionality reduction, and LDA topic model is used in the research. At the same time, due to the coarse granularity of the topic extraction in the LDA algorithm, it is not suitable for short text, this paper proposes a new topic model based on improved k-means and TextRank and gets good results on this dataset. This paper uses multiple data mining models to analyse film reviews and presents an empirical study of the efficacy of machine learning techniques in text semantic mining.

Aiming at the evaluation of equipment competitive acquisition management, the paper proposes an evaluation method based on entropy theory. First, the concept of equipment competitive acquisition management entropy is defined. Second, the conceptual model, mathematical model, and evaluation index system of equipment competitive acquisition management entropy are constructed. Finally, taking policy and regulatory entropy as an example, the paper has calculated and analyzed its evaluation results to prove the effectiveness of the method.

Big data technology provides convenience for all entities in the supply chain to obtain demand forecasts and share the information. This article considers a supply chain composed of a manufacturer and two competitive retailers and analyzes the value of sharing demand information in the supply chain. In this supply chain, the manufacturer has a hybrid MTS/MTO production system and sells products to the MTS retailer and the MTO retailer. Both the manufacturer and the retailers have private demand information. We established a no-information sharing model, a full information-sharing model, and two partial-information sharing models, to study the value of sharing information. The results show that the full information sharing strategy cannot benefit all entities. However, if the demand forecasts of the two retailers are very different and lower than the manufacturer's forecast, sharing information between the manufacturer and the retailer who has high demand prediction can benefit all entities.

Unmanned Aerial Vehicles(UAV) swarm is a rapidly developing field, and with it comes the need to identify the swarm based on observations. The problem of trajectory clustering is put forward in the identification of UAV swarms, especially modularized UAV swarms. We propose a new method of Network Integrated trajectory clustering(NIT) to solve the trajectory clustering problem in a fast-changing and chaotic environment which requires a quick response, fault tolerance, and accuracy. The experiment results prove the flexibility and adaptability of the NIT method towards various demands and multi-dimensional data. Moreover, the algorithm proposed based on the method shows priority over the other three trajectory clustering methods(DTW, Fréchet distance, GMM) on the accuracy, and fault tolerance in clustering swarm trajectories. The method raised in this paper is an innovation to both multi-agent systems identification and trajectory clustering methods.

Multi objective optimization problem (MOP) usually has more than two objective functions, and the optimal solution based on Pareto frontier is obtained. The traditional optimization algorithm cannot meet the needs of industrial application when dealing with multi-objective optimization problems. With the good performance of evolutionary algorithm in solving complex problems, its application field is also extended to multi-objective optimization problems. The Pareto optimal solution and evaluation system of multi-objective optimization problem are analysed. Particle swarm optimization (PSO), one of the evolutionary algorithms based on swarm intelligence, is briefly introduced. The combination of particle swarm optimization algorithm and multi-objective optimization is studied.

The paper focuses on children's data present inElectronic Health Records(EHRs) starting fromraw data gathered from Southwest Medical Data Center in China. We designed a systembased on workflow to analyzefeatures that affect children's growth. The system is able to relate health caredata to diagnosis codes and with additional information integrated and correlated to EHRs data. Finally, we propose prediction models based on Recursive Feature Elimination (RFE), which can identify features thatare important to detect children's growthand findcorrelations among features.

The current mobile learning service platforms are mostly self-centered, and they provide exactly the same learning content, and their personalized service capabilities are insufficient to meet the different learning needs of different students at the same time. To this end, combined with data mining technology, the research on a mobile learning service platform that can provide personalized services is conducted in this paper. The platform design includes four parts, namely frame design (development frame design and network architecture), main hardware design (microprocessor, switch, memory), main software design (user registration and login program, personalized mining analysis program, online learning procedures, online communication procedures) and system testing. The test results show that the concurrent performance and compatibility performance of the system meets the requirements after testing, allowing multiple users to use different mobile clients to access the platform at the same time.

In the analysis of e-commerce suppliers, the use of big data analysis technology is very important. The interference factors in the analysis process determine the supplier's choice. According to the characteristics of e-commerce, the e-commerce competition model, evaluation matrix, and the intelligentized degree of each index are determined through level analysis. Experience has shown that accurate data on the financial status of the organization can be obtained through intelligent screening methods, which has laid a solid foundation for improving the good cooperation between the original electronic companies and suppliers.

Big data and artificial intelligence technology, as the unified means and carrier for collecting, storing and computing core data of contemporary power IoT, has diverse and complex characteristics of its data sources and types. And the lack of insecure data access, abnormal response and terminal access rights control leads to the failure of the integrity and credibility of the closed-loop information defense. In particular, wireless sensor networks (WSNs) in power scenarios are more susceptible to malicious attacks. To address the above problems, this paper proposes a power IOT information defense strategy based on improved identity-based dynamic clustering authentication algorithm (IIDC). First, the terminal device sets the private key to solve the key escrow problem of terminal security authentication in the IOT model. At the same time, the improved algorithm dynamically generates pseudo-cryptographic matrix to avoid collusion attack. Finally, a hierarchical privilege management mechanism is adopted to decrypt the terminal authentication once. The proposed algorithm is more suitable for terminal security access and power consumption requirements in WSN-based power IOT scenario, as verified by experimental simulation.

K-M (Keenan motley) model calculates the total loss value of indoor wireless signal penetrating multiple walls by adding fixed loss values of single wall, without considering the influence of wireless signal polarization and incident angle on wall loss. In order to solve this problem, considering the polarization and incidence angle of the beam, the wall and its direction can be detected automatically from the blueprint image by Hough transform in image processing. By obtaining the direction and angle of the wall, the loss attenuation is defined as a function of the incident angle. The improved model can define various types of walls with different attenuation characteristics. Through the comparison of actual measurement, the error between the improved model and the measured road strength loss is reduced, and the root mean square error between the measured value and the prediction value of the improved model is obviously improved, and the prediction accuracy of the model in indoor environment is improved.

In the research of solar power prediction, providing accurate prediction data in real time is one of the most effective means to enhance the capacity of wind power acceptance and improve the power reliability and economy. The existing prediction models based on statistical methods are often unavoidable in data preprocessing and model training stage, and their adaptive ability needs to be improved. Considering that the sparse coding method does not require model training, and has the characteristics of high solving efficiency and strong self-adaptability, an online solar energy prediction model using sparse coding is proposed.Firstly, the historical time series data is composed of input-output pairs with delay, and the dictionary is respectively constructed in atomic form. Then, the sparse weight is calculated for the delay input data vector to be predicted, and the corresponding predicted output is obtained by borrowing the dictionary. Taking the actual solar power data of Alberta, Canada as sample, the simulation was carried out in MATLAB. The simulation results show that the model can accurately predict the solar power and improve the effectiveness and practicability of the prediction

With the rapid development of power system, the requirement of network clock synchronization accuracy is becoming higher and higher, in many cases, it needs to reach microsecond level. In order to meet the increasing demand, based on the analysis of the basic principle of PTP clock synchronization protocol, a physical layer clock synchronization system based on STM32 MCU and DP83640 chip is proposed. The system synchronization time error is less than 1 microsecond by using the timeacc-007 time precision measuring instrument. The scheme meets the requirements of accurate clock synchronization for all distributed nodes in the power system.

Coronavirus disease is seriously affecting the world in 2019. Wearing a mask in public places is a major way to protect people. However, there are few studies on mask detection based on image analysis. In this paper, an improved mask wearing inspection algorithm based on the SSD algorithm is proposed. The SSD algorithm is improved to add a face mask wearing detection task. Based on the original SSD model, the algorithm improves the mask wearing detection capability by introducing inverse convolution and feature fusion in combination with an attention mechanism to filter out the information to be retained. A dataset containing 3656 tensor images was created and manually labeled for network training. Experiments on this dataset show that the algorithm has good accuracy for mask wearing inspection.

The traditional system takes a long time to adjust the fresh food storage temperature. Therefore, under the background of "village revitalization", the comprehensive service system for fresh food e-commerce is designed in this paper. First of all, on the hardware side, by optimizing the physical architecture, the application layer and the data server layer are separated; on the software side, the system database is designed to extract customer order data. Then the ant algorithm is used to select the optimal logistics configuration path, and the temperature control algorithm is used to adjust the temperature environment so that the fresh food reaches the optimal storage temperature in the transportation path. Finally, the hardware and software design are combined to realize the comprehensive service system design. The experimental results show that compared with the traditional system, the designed system in this paper shortens the control time of fresh storage temperature and strengthens the service control of cold chain logistics.

Many applications of wireless sensor networks (WSN) require nodes to know their location information. Monitoring data without node location information is meaningless in many situations. Since the node positioning efficiency of WSN directly affects the performance of the network, designing an efficient positioning algorithm is the focus of WSN research. Based on the principle of TDOA positioning in WSN, it studies two main positioning algorithms: Chan and Taylor algorithm. Meanwhile the least square (LS) method is used to provide the initial position for Taylor algorithm and the first calculation result for Chan algorithm. Simulation results under different conditions show that the Chan algorithm has a good positioning accuracy and less computation than that of Taylor algorithm in the line of sight (LOS) environment. However, under the actual channel, due to the influence of No Line of Sight (NLOS), the positioning accuracy of Taylor algorithm is better than that of Chan algorithm, but its computation is large.

Aiming at the problems that, it is difficult for the administrator to obtain various state information in the forest in real time, and it is impossible to make real-time and accurate judgments on the behaviors that endanger the forest and formulate effective countermeasures, a forest security guard service system based on Beidou positioning is designed. The system adopts STM32F103C8T6as the main control chip to control different types of sensors such as flame sensor, temperature sensor, humidity sensor, sound sensor and so on, the information of temperature, humidity, illumination, fire situation, animal type, personnel alarmaround the nodes in the forest environment associated with the location information of the Beidou module is transmitted to the server through the GPRS network in real-time, then the server broadcasts the normal information and the alarm information to the administrator through the control panel and the mobile client. The system adopts solar energy and can monitor the forest environment information automatically, which can greatly reduce the workload of the employees, low the costs of the network's layout and monitoring, and provide the researchers with massive and reliable research data. It has a wide range of application in the forest management, agricultural monitoring, environmental monitoring and other fields.

The protocols in the Internet of Things are more complex than the Internet. Once the Internet of Things devices have security problems, it will lead to a decline in user experience and huge economic losses. Therefore, this paper proposes a secure communication scheme for the Internet of Things based on a dynamic key update mechanism. This paper uses a variety of chaotic equations with good random characteristics to generate the session key library, and then uses the lightweight symmetric encryption algorithm Present-80 to encrypt the communication data. The simulation test results show that under the condition of ensuring communication efficiency, the security communication protocol of the Internet of Things based on the dynamic key update mechanism proposed in this paper has higher security.

At this stage, due to the increasing use of electric vehicles, the position of electric vehicle load scheduling in grid power scheduling is becoming more and more important. Effective electric vehicle power dispatching can balance the peak-valley difference of power dispatching, increase the power supply utilization rate of power grid dispatching, and reduce the power supply pressure of line transformer. The load forecast can describe the user's electricity consumption habits in the next period of time, and can provide important data basis for power dispatching. This paper summarizes the research status of electric vehicle charging load, analyzes traditional charging load research methods, propose a charging load forecasting method combining XGBoost(Extreme Gradient Boosting) and LSTM (Long Short Term Memory Network), And use the data of a charging station in Jiangsu to verify the calculation example. The proposed method is based on the prediction results of the XGBoost model for feature engineering, extracting data features using phase space reconstruction techniques and statistical methods. In addition, training the LSTM model for load prediction. Based on the charging record data of domestic charging stations, this paper applies the artificial intelligence method to the charging load forecast of domestic charging stations for the first time. The charging station load forecasting method studied in this paper can support the regional load forecasting research of electric vehicles with high permeability, and further optimize power dispatching.

As the first step of cybersecurity situational awareness, the accuracy of cybersecurity element recognition will directly affect the results of situational understanding and situational prediction. In this paper, we propose a network element recognition method based on the convolutional attention mechanism combined with a long- and short-term memory network. The input network traffic data is successively passed through the convolutional neural network, attention mechanism, and long- and short-term memory network, which not only takes into account the influence degree of different network attributes on different network behaviors but also realizes that the feature information extracted in the early stage can be circulated in the network, thus providing a discriminant basis for the final network behaviors To verify the effectiveness of our proposed method, we perform experimental validation on the KDD-Cup 1999 (kdd-99) dataset. The results show that our proposed method achieves an accuracy of 98.48% in the identification of network security elements. In addition to this, we also compare and analyze our proposed algorithm with other mainstream algorithms, and the results also validate the effectiveness of our proposed method.

In a society with highly developed information and communication capabilities in the 5G era, the traditional way of thinking has been unable to meet the digitalization requirements of efficient and feasible. Therefore, we propose a digitalization method to solve practical problems, which is called programmed thinking. In this way, we can imagine each problem as a combination of multiple functions, so that these functions can be determined according to the overall rationality of the requirements of the data content exchange, as well as the overall scheme design. Because this method is implemented in the form of modularization through the overall requirement analysis, even if there is an exception, it is also no need to redesign the entire requirement or function, only need to replace and modify the abnormal part of the method function, so as to achieve the effect of rapid response, switch at any time and simple upgrade.

In order to improve the accuracy and efficiency of the vehicle network security situation prediction, a prediction algorithm based on improved CLPSO-RBF is proposed. Firstly, for speeding up the optimization efficiency of CLPSO, a reasonable speed monitoring variable has been introduced. Secondly, we use the improved CLPSO algorithm to optimize the clustering radius of the RBF neural network, so that the optimal RBF network structure can be determined. Finally, we use the optimal RBF network to predict the security situation of the vehicle network. Simulation experiments have proved that the improved algorithm has higher accuracy and faster convergence rate in situation prediction, and has better prediction effects.

Security model is the main means to protect the network information security of vehicle. Due to the rapid development of artificial intelligence in recent years, machine learning technology is also emerging in the field of Internet of vehicles security. The random forest model is a strong classifier and can prevent overfitting better than the decision tree model. However, only using the traditional random forest invasion detection model has some problems, such as: the model detection time is long, the false alarm rate is high, the ability of using platform transplantation is poor, etc. In this paper, it is optimized in a lightweight way to reduce the time consumption and improve the accuracy of intrusion detection in the vehicle networking intrusion detection model.

The quality and safety of fruits, vegetables, and other agricultural products is the lifeline of agricultural development. In recent years, the state departments concerned have strengthened the whole process block on the quality of fruits and vegetables, continuously optimized the measures, and continuously improved the quality control level of fruits and vegetable products. Traceability code is the key technology to realize the whole process block and traceability of the whole industry chain. There is a risk that the traceability code will be modified privately in the application process. Once the traceability code is cracked and modified, its important role in the traceability system will be destroyed. In order to prevent the traceability code from being arbitrarily modified, this paper designs and studies an optimization algorithm of traceability encryption based on an embedded system

The early-warning and inspection system for the intelligent greenhouse is an inspection robot based on the Internet of things and equipped with various sensors and intelligent cameras with 4G or 5G communication functions, such as temperature and humidity, light, gas, and flame. Which uses the wireless network transmission to the cloud computing platform for data analysis, sending warning information through smartphones to enable farmers to real-time monitoring in agricultural greenhouse crop growth status, temperature and humidity, illumination, CO ₂concentration, plant diseases, and insect pests, etc., to achieve automatic control, scientific planting and raising the yield and increase their income. This paper mainly analyzes the hardware and software structure of this early-warning and inspection system, data transmission and processing, and Android application, so as to promote the application of intelligent planting.

With the outbreak of novel coronavirus pneumonia worldwide, a kind of epidemic prevention robot is urgently needed in society. The intelligent temperature measurement robot system mainly includes the robot's high stability multi-attitude omnidirectional movement module, robot path planning and autonomous navigation module, robot environment information acquisition and recognition module, robot remote control module, and robot system integration. According to the design principle of the wheeled robot, high stability and multi attitude omni-directional mobile platform of temperature measuring robot are constructed; path planning and autonomous navigation of robot are realized based on sensor data fusion; temperature detection and remote care function are realized by thermal imaging sensor system by capturing the thermal radiation emitted by the human body, and based on open modular control system architecture. Based on the existing research foundation, the intelligent temperature measurement robot system is established by integrating the robot multi attitude omni-directional mobile system, robot path planning, and autonomous navigation system, robot environment information acquisition and recognition system, and robot remote control system.

Periodic maintenance of transmission lines is a necessary guarantee for stable power transmission and avoiding serious power accidents. In order to reduce the safety risk of the first person climbing the transmission tower and the last person going down the tower when climbing the transmission tower manually, this paper develops an auxiliary climbing robot for the maintenance of transmission tower based on the bionic climbing principle. It includes the modular mechanical structure of the main body lifting, obstacle crossing telescopic, terminal clamping, and body jacking, as well as airborne motion control system and hierarchical control system of upper computer monitoring terminal. The comprehensive test results of indoor and outdoor simulated transmission tower environment show that the robot system can effectively complete the high-altitude climbing of power tower, and ensure the follow-up tasks of hanging, dismantling, and maintenance of safety anti-falling device.

As the underlying platform for security applications of the Internet of Things, blockchain can provide reliable storage, data encryption, access control, identity authentication, and other service support for the security of the Internet of Things. Its immutability can ensure the authenticity of the information on the chain and the construction of shared accounts The foundation of trust is to ensure business transparency and traceability. Asymmetric encryption can ensure the information security of the Internet of Things. Smart contracts support automatic transactions of IoT devices to improve the operating efficiency of the system, thereby promoting the flow of IoT data and enabling timely warning and carry out effective control. The application research and industrial application of blockchain in the energy field will bring innovative application models to the Energy Internet and realize the flow of data value. This article focuses on the energy consumption monitoring of the Energy Internet consumption, applies blockchain technology to the construction of an energy consumption monitoring system for key energy-consuming units, analyzes the fit between business needs and blockchain technology, designs a blockchain platform architecture, and discussed the issues of energy consumption monitoring blockchain terminal equipment authentication and collection of trusted data.

The existing wireless network multi-channel allocation technology has the defects of low throughput and large average delay. Therefore, this paper proposes a multi-channel allocation technology for large-scale connection scenarios. Through the wireless network architecture, the large-scale connection scenario is described. Based on this, the interference model of wireless network is built. In order to simplify the process of multi-channel allocation in wireless network, Kruskal algorithm is used to form the minimum spanning tree of wireless network topology, and the link load weight is calculated. Based on the calculation results, the wireless network multi-channel is allocated according to the link load weight from large to small The multi-channel allocation of wireless network for large-scale connection scenarios is realized. Using the proposed technology to determine the value of link load weight, the multi-channel allocation scheme is obtained, and the simulation experiment is carried out. The experimental results show that: compared with the standard value, the proposed technology has large throughput and small average delay, which fully shows that the proposed technology has higher allocation performance.

In recent years, multi-agent reinforcement learning (MARL) has been applied widely, especially in large multi-role games such as StarCraft and unmanned aerial vehicles (UAVs) combat simulations. However, MARL is faced with challenges regarding fast convergence and efficient cooperation. In a multi-agent scenario, on the one hand, when a fully centralized network model is adopted, it is difficult for the model to converge due to the huge action space; on the other hand, it is difficult for a decentralized model to cooperate and achieve global optimization. To jointly control multiple agents, we propose an improved PPO algorithm by combining a centralized network and decentralized networks. Our method not only reduces the action space and accelerates the convergence, but also introduces more diversity for agents' decision-making.

With the rapid development of artificial intelligence and the Internet of Things technologies, the degree of agricultural Informatization and intelligence has been further improved. Governments of every country attach great importance to agricultural modernization and intelligence. Based on this purpose, this project completely independently designed an intelligent mobile picking robot for application in agriculture. The picking mobile robot is mainly based on the STM32 development board, which is loaded with the color sensor, infrared sensor, and ultrasonic sensors, which is instead of the CCD sensor to identify maturity, thereby reducing the difficulty of the algorithm and realizing fast picking tasks. In the harvest season, the robot will be mainly used for picking fruits, which can completely, replaces farmers' operations, liberating labor, and breaking the constraints of time and space.

With the development of Internet of Things technology, it drives the new development of most industries. Combining the fermentation industry with the LOT, fermentation is a highly nonlinear process in which operators need to measure and operate regularly. This paper proposes an efficient growth and metabolism control method for lactic acid fermentation based on the Internet of Things, using Internet of Things technology combined with cloud computing. Through neural network and expert system of predicting, the fermentation process can be controlled more accurately and effectively, and the threshold value can be changed in time according to the growth stage of the microorganism, to a certain extent, to avoid the error of manual operation. Show the advantages of combining the Internet of Things.

Because of the damage in its major parts such as the plunge pool and sluice gates, the Kariba Dam is now in a dangerous state that needs urgent actions of rehabilitation. In this paper, we use the multiple dam system in Kariba to verify the dam performance models we propose. First, a dam performance model is built with 3 parts which are focused on hydropower generation, water storage capacity, and the costs of a single dam respectively. Next, we simulated the geometrical characteristic of the dams based on the terrain of Zambezi river basin. The number and placements of the multiple dams system are determined to satisfy the generation and water capacity of the existing dam. Discussion of the costs-benefits difference between multiple dams system and other 2 solutions are made, the result shows that mulitple dam system is the optimal solution, and the model we proposed were proved to be reliable.

Aiming at the problem of fuzzy C-means (FCM) in the estimation of underdetermined mixing matrix, that the estimation accuracy is not high and the robustness is poor, a density peak clustering (DPC) based on density peak clustering (DPC) is proposed. Improved Kernel-based Fuzzy C-means (KFCM). The kernel function is introduced into the FCM algorithm to construct the KFCM algorithm based on the Gaussian kernel function, which can effectively overcome the influence of noise points and isolated points on the clustering results and improve the estimation accuracy of the mixed matrix; the traditional DPC algorithm is improved and merged with the KFCM algorithm, and thresholds are set for the local density and high-density distance to achieve the initial clustering center of the KFCM algorithm and the automatic determination of the number of cluster centers improves the robustness of the algorithm. Experimental results show that the algorithm has greatly improved the estimation accuracy and robustness of the underdetermined mixed matrix.

With the promotion of "Internet +" and smart cities, as an important data voucher, electronic certificate brings great convenience to people's life. However, the phenomenon of "data island" based on the electronic certificate database of each city hinders the sharing of an electronic certificate. In addition, problems such as centralization of data storage, poor security, and tamper-proof modification are common in electronic certificate libraries. To solve these problems, we present a blockchain-based electronic certificate sharing scheme, which uses the InterPlanetary File System (IPFS) and Ethereum smart contract to achieve secure storage and data sharing of electronic certificate by taking advantage of the non-tampering and decentralization of blockchain. At the same time, we achieve fine-grained access control by using attribute-based encryption of ciphertext policy and assigning attribute keys to users without affecting retrieval efficiency. Finally, through the simulation and performance analysis of the Ethereum test network, the analysis results show that our scheme is effective and feasible.

Hidden services are a feature of Tor(The Onion Router)[1]. It provides anonymity for the service requester while maintaining the anonymity of the service provider. Since it is quite difficult to trace back and locate both parties in the communication, the criminals use hidden services mechanisms to construct various illegal activities in the darknet, which has brought adverse effects to society. In order to prevent the abuse of Tor hidden services, the discovery and analysis of hidden services are particularly important. The aim of this survey paper is to review and compare the literature of the past five years, provide the readers with methods for discovering tor hidden services, along with the various content analysis methods developed and proposed from time to time. we explain their key ideas and show their interrelations.

In order to improve the hit rate of memory pages and prolong the service life of flash memo ry, an adaptive double area page replacement algorithm is proposed. The buffer is divided into page area and page cluster area. The initial size o f each area is set. The size of the area is dynamically changed according to the number of page hits in the two areas. The replacement unit of the page area is a single page. The CCF-LRU algorithm is used in page area replacement policy. When the page in the page area needs to be replaced, the rep laced page is put into the page cluster area. The replacement unit of page cluster area is the whole page cluster. When the page area cluster needs to be replaced, the dirty memory pages in the page cluster are divided into small memo ry pages with the same size as the flash memory pages, and only the dirty small memory pages are written back. According to the analysis of experimental results, compared with CF-LRU, CCF-LRU and FAB, the hit rate is increased by 9%, 8% and 11% respectively

In order to solve the problems of unbalanced energy consumption and shortened network lifetime caused by the uneven distribution of cluster heads (CHs)in wireless sensor networks, a Limited IST and Random Uniform Clustering Algorithm (LIRUCA) is proposed. The algorithm comprehensively considers the energy of nodes, the distance to the sink, and the correlation between nodes when selecting CHs, and limits the intra-cluster communication distance, thus improving the uniformity of CH distribution. Compared with LEACH, EBCRP, and LEACH-improved protocols, results of simulation experiments show that the LIRUCA algorithm can effectively improve the uniformity of clustering, reduce the energy consumption, and prolong the network lifetime.

With the increasing maturity of WebGL technology as well as the development and construction of the 3D geographic information system, 3D geographic information sharing service has become the trend of future development. There are numerous three-dimensional data sources of the power grid and many applications across departments and systems, but the three-dimensional geographic information service based on unified standards is still rarely seen. The paper studies and discusses the selection of 3D service standards and the implementation technology of 3D geographic information service for the power grid.

Aiming at the low security and low real-time performance of the message authentication security layer (MASL) message authentication code (MAC) algorithm in the railway signal security communication protocol RSS-II, the advanced encryption standard AES is used as the core algorithm of the MAC code and the ciphertext is used Packet link encryption and decryption of the message, simulation of the security protocol from the security-related key service process of the RSSP-II communication protocol, verification of the confidentiality of the improved protocol, and optimization of the protocol key service process, that is, the key The center uses the advanced encryption standard AES algorithm to authenticate the vehicle equipment and the ground wireless block center, further strengthening the security of the protocol.

In order to solve the problem of repeated collection of ownership information in the process of real estate integration confirmation and registration, a collection system applet was specially developed, which changed from the original technical staff from door to door collecting ownership information to government-led and farmer cooperation. According to the simple and clear system interface, the farmers enter the ownership information, upload the photos of the right source materials, the homestead and the house photos, which solves the big problem that the ownership investigation must be repeated and the ownership information must be entered one by one, which has changed to a certain extent. The operation mode of traditional ownership investigation has greatly reduced the pressure of field work, reduced the cost of manpower and material resources, improved work efficiency, and protected the legitimate rights and interests of farmers.

With the rapid development of social humanities and technology, especially the development of the Internet, the current Internet era, the amount of information is increasing rapidly with explosive growth rate, the information overload problem is particularly obvious, on the accurate acquisition of information from the massive amount of information users want, from which the personalized recommendation technology was born. In order to solve the problem of acquiring the information users want, this paper researches and discusses a kind of personalized recommendation algorithm - a user-based collaborative filtering algorithm, analyzing the user behavior, comparing the advantages and disadvantages of other related algorithms, using the UserCF algorithm, and optimizing the sparse matrix to reduce the time and complexity of the operation. The algorithm is implemented by software to generate recommendation results. The results of the experimental data in this paper show that the algorithm is effective in recommending projects to users.

With the development of vehicular Ad hoc Networks (VANENTs) technology, application in VANENTs has becomes an important role in human life. In practice, the location information of not all vehicles in VANENTs contains sensitive information. But traditional publish algorithm for geospatial stream for VANENTs based on differential privacy assumes that all the data in the geospatial stream contain sensitive information of users. As a consequence, utility of the traditional algorithm has been decreased. In this paper, a novel algorithm is proposed to publish geospatial stream, denoted as PGSV. PGSV make used of one-sided differential privacy to guarantee sensitive information. In the end, we make used of public dataset to verify efficiency of PGSV. The result show that PGSV have increased utility of published result and guaranteed security of published result.

To manage the electric network effectively, 3D visualization framework is necessary to reveal the complexity of power grid system. In this paper, an interactive online 3D visualization framework is proposed to visualize the electric network related information from power station to the static data of the system. We applied Cesium, an open source platform for 3D visualization based on WebGL supported browsers. Meanwhile, multiple Level of Detail (LoD) structure of power grid facility is generated to improve the speed of 3D model loading and rendering. The experimental results indicate that the proposed framework can integrate and dynamic visualize power grid information from multiple sources. The multiple LoD structure can reduce the model loading time by 76% and increase the FPS over 50% according to the tests on sample data.

In recent years, botnets have used the domain generation algorithm to generate dynamic typified malicious domain names to bypass various detection methods. Given the depth detection model of such domain names, domain names are generally processed by filling and transforming them into a fixed-length one-dimensional vector and then classifying them with poor detection performance. Therefore, this study first divides the domain into a word array and converts it into a word vector using pre-trained word vector models, Embeddings from Language Models. The domain is inputted into the TextCNN model for training classification. From approximately 100,000 data sets, a 94.22% accuracy rate and 6.87% FPR value can be obtained from the training. Compared with previous detection models (i.e., LSTM and CNN), more training and testing are needed, but improvements are made in all indicators.

With the rapid development of the Internet, sensor network, and mobile Internet technology, a large number of data sets are continuously generated in the form of streaming in various application fields. At the same time, the processing of stream data has drawn more and more attention because of its application in different situations. To meet the urgent need for the processing of streaming data, there are many computing engines on the market, such as Spark and Flink. Traditional data-flow analysis has many problems in the process of streaming data, such as the high time delay, weak extensibility, and bad adaptability. In order to improve them, people use Kafka as the intermediate-cache and the computing engine that uses Spark Structured Streaming as the streaming data to realize the effective process of multichannel flow data. Considering the present demand for low latency and high flux that are analyzed by urban management in traffic video, processing the data of multichannel traffic video on distributed computing platform realized the tracking and search on motor vehicles, non-motor vehicles, and pedestrians of multiple roads and intersections.

The orbital angular momentum (OAM) of light has been considered as a promising degree of freedom (DoF) that gives access to a higher-dimensional Hilbert space, which may lead to potential higher capacity quantum communications. Due to the fragility of the OAM state, the traditional view is that turbulence will make OAM-QKD infeasible in satellite-to-ground channels. However, based on the detailed phase screen simulations of the expected atmospheric turbulence, we find that quantum key distribution (QKD) using OAM of the light is feasible in certain system configurations, especially if quantum channel in-formation is utilized in the processing of post-selected states. Therefore, we propose a satellite-to-ground quantum key distribution protocol based on the orbital angular momentum of the light, which uses the principle that OAM-QKD can only be used in high-altitude ground stations with larger receiver apertures with-out using classic optical probes. At the same time, the classically entangled light is used as a probe of the quantum channel and reasonably-sized transmitter-receiver apertures are also employed. Numerical simulation results show that this protocol can lead to positive secret key rates even under circumstances where a sea-level ground station with a reasonable-sized aperture is used. We also found that quantum channel conjugation enables a key rate advantage provided by the higher dimensions of the protocol to be realized.

The development of 5G communication and Internet of Things (IoT) technology has brought massive data transmission. This paper considers a first-in-first-out packet data transmission scenario with actual circuit power consumption and proposes a data transmission strategy with optimal energy efficiency. Specifically, the proposed algorithm in this paper uses a double-layer iteration algorithm, then compare and replace the optimal solution of the ideal circuit power consumption with the optimal solution of the actual circuit power consumption. The numerical results show that our proposed algorithm can significantly reduce energy consumption.

In the satellite-ground quantum key distribution network with power IoT, due to the imperfection of the quantum equipment and the noise in the channel, the raw key will have some bit error rate and cannot be used for secure communications. To solve this problem, we propose an improved data reconciliation model based on Turbo for satellite-ground quantum key distribution to extract the raw key, and then the key on both sides of communication are identical. We carry on the optimization and simulation to Turbo and increases its error correction efficiency by 15%. To test and verify the feasibility of the model, we implement satellite-ground quantum key distribution simulation experiments and respectively use the low-density parity check (LDPC), Cascade, and Turbo for data reconciliation. Experimental results show that Turbo can make the final key generation increased by 8% comparing with LDPC and Cascade.

In order to solve the problem of large fluctuations in UWB(Ultra Wide Band) positioning accuracy and solid-state deviations in the positioning results, which are in complex indoor environments, an adaptive Kalman filter method is introduced in the later stage of data processing. This method can better retain data information and obtain better filtering effects when the system noise is complex and measurement information is missing,.Thereby providing better positioning accuracy. The experiment uses AGV trolley equipped with UWB sensors to obtain measurement data, uses traditional Kalman filtering and adaptive Kalman filtering methods to compare the filtering effects of the algorithms. The experimental results show that when the measurement information is missing, compared with the traditional Kalman filter algorithm, the adaptive Kalman filter method is a real-time high-precision indoor positioning algorithm with higher positioning accuracy.

In the process of data interaction, neighborhood query, filtering, visualization, and dynamic update of LiDAR point cloud data, how to efficiently organize and process massive point cloud data, and quickly index and locate any point in the point cloud and its neighborhood Search is a key issue to be solved urgently. In this paper, combining the advantages of a virtual grid with no interpolation loss on original data, sT spatial relationship, and low memory occupation of the octree, we design an index method based on the combination of virtual grid and adaptive octree based on dynamic scheduling of internal and external memory. Realize the organization and scheduling of massive LiDAR laser scanning point cloud data.

In response to the call of strengthening the construction of the ubiquitous electric power Internet of Things by the State Grid Corporation, combining the urgent requirement in the construction of smart cities, this paper researched the application of wireless communication technology on smart street lamps based enhanced machine type communication (eMTC) technology. Based on this research, a kind of single-lamp controller PEC6403 based on eMTC wireless communication technology was developed. In order to realize remote control of the lamp, run data sampling in real-time and malfunction alarm, the TCP long connection established by the remote management platform and the eMTC wireless communication module was used to receive and send a message by this single-lamp controller. At last, this paper did some experiments to certificate the high communication rate, low latency, and low power consumption of this eMTC wireless communication module. Finally, this paper analyzed the problems existing in the large-scale promotion in the field of electricity application of the product and put forward the prospect of eMTC technology application in the electric field in the future.

A wireless electrical stimulation device based on biological electrode array is developed. The stimulator uses the FPGA as the central controller, a 4 × 4 electrode array is built and direct digital synthesizer (DDS) is used to generate stimulating waveforms. To generate adequate current load on human body, an isolated bi-direction constant current source module is designed which can detect electrode loss, as well as provide safe bi-direction constant current. The host computer uses an interactive software for android system, and the stimulating amplitude, pulse width and frequency and the electrode selection can be adjusted flexibly. The ES electrode arrays device has the characteristics of small size, good stability, low power consumption, easy to operate. It will bring convenience to clinical applications and bioelectrical stimulation scientific researches.

Mastering the laws of wildlife activities and family territory is an important task for wildlife protection, habitat restoration, and epidemic surveillance. Aiming at the shortcomings of traditional wildlife monitoring methods, an overall framework of wild animal satellite tracking system is proposed, and key technologies such as MEMS design, energy optimization, miniaturization and lightweight, wearable methods, and big data platform development are discussed and practiced. The results show that the wild animal satellite tracker developed by the above technology has the characteristics of miniaturization and lightweight, which can better solve the problem of energy supply in the field, and the service life can reach more than 5 years. Cloud recording of wild animal tracking data, Cloud Storage, Cloud computing, and Cloud monitoring. It plays a role in ensuring the data security of China's wildlife ecological resources and other sensitive geographic information, and also provides scientific and effective technology tools and data support for China's wildlife research, protection, and management, and is of great significance for serving the national ecological civilization construction.

In recent years NUMA architecture based on multicore system leads a mainstream solution to tackle the rapidly expanding computation in bioinformatics analysis. Sequence alignment is the most computing cost in the workflow of variant analysis in bioinformatics. Although parallel coding in sequence alignment could be few difficulties, approaching a good performance in such a system is complicated. This research explores the architecture bottlenecks we usually neglect in the implementation of bioinformatics computing, NUMA effect for instance. We exploit the scalability of threads in the sequence aligners to illustrate the problem and the significance of NUMA architecture in the multicore system. The preliminary results in the experiments display that sequence aligners cannot take advantage of NUMA-based multi-core architecture. The scalability of threads is deficient, even negative in time-cost at the case of large-scale genome data larger NUMA nodes.

Collaboration log, which can be used to reflect the trust of an organization and applied to subsequent analysis and audit, is constantly generated during the organization's collaboration. Common systems usually store them by a third-party organization, which exists at risk of tampering. Regarding this issue, blockchain can safely manage the collaboration logs owing to its decentralized and unforgeable features. In this paper, we implement an organization collaboration prototype system for spare parts procurement based on the Software As A Service(SaaS) collaboration model and Hyperledger Fabric solutions. To validate the performance of the prototype system, we conduct some performance tests on the prototype system. The result shows the feasibility of our proposed system.

The honeypot system is a critical defense facility in cybersecurity. It captures and analyzes the attack behavior for constructing a more robust defense system. However, considering the overhead caused by analysis, the system cannot support fine-grained analysis. This paper proposes RRPOT, a novel honeypot system based on the record and replay (RnR) framework of virtual machine (VM). By combining the native honeypot system with an RnR framework, the system supports to postpone the time-consuming analysis to the replay stage, thereby enables fine-grained analysis. Moreover, we propose an on-demand record mechanism to reduce performance and storage costs during the record stage and employ an on-demand replay mechanism to improve replay efficiency. Leveraging these approaches, we develop a flexible and efficient honeypot system, and it was proven useful in the practical scenario.

Aiming at the problems of low efficiency of path planning, a new node storage structure is introduced and the search method is optimized to improve the deterministic iterative path planning algorithm. First, the number of antibodies is determined based on the connectable starting path point, when generating the initial antibody with the inspiration of the optimal angle vaccine. Then, the connectable path point of the starting point is treated as the root node to rebuild the new path with path filtering by means of the optimal path fit value. The initial optimal antibody is used as the screening criterion to avoid invalid path point mutation with 100% confidence conditions. Finally, a path point is a basic unit that forms the connection network and stores the parameter information that reaches it. After the algorithm iteration, the optimal pathis output. In different maps, the algorithm in this paper is compared with others. The results show that the algorithm effectively reduces the computational cost and has better adaptability to different maps.

Kalman filter algorithm, an effective data processing algorithm, has been widely used in space monitoring, wireless communications, tracking systems, the financial industry, and so on. On the Sunway TaihuLight platform, we present an improved Kalman filter parallel algorithm which is according to the new architecture of the SW26010 many-core processors (260 cores) and new programming mode (master and slave heterogeneous collaboration mode). Furthermore, we propose a pipelined parallel mode for the KF algorithm based on a seven-level pipeline of the SW26010 processor. The vector optimization strategy and double buffering mechanisms are provided to improve the parallel efficiency of Kalman filter parallel algorithm on SW26010 processors. The vector optimization strategy can improve data concurrency in parallel computing. In addition, the communication time can be hidden by double buffering mechanisms of SW26010 processors. The experimental results show that the performance and scalability of the parallel Kalman filter algorithm based on SW26010 are greatly improved compared with the CPU algorithm for five different data sets, and is also improved compared to the algorithm on GPU.

The node differential privacy (node-DP) can be used to protect the private information of nodes and edges in the graph. In this paper, we propose an algorithm of publishing node strength histogram under node-DP, which could improve the accuracy of publishing the node strength histogram. In this algorithm, we use the Sequence Edge removal to reduce the sensitivity of query function and restrict the weight of edges to make the distribution of node strength denser. Furthermore, we use the histogram grouping algorithm Hierarchical Cluster Grouping to group the buckets to merge buckets with close values into one group. The experiments show that our algorithm maintains higher data utility than those traditional histogram publishing algorithms under the same privacy budget.

Three-tank water tank is one of the most extensively used and representative industrial mathematical models, which has very important research significance. Liquid level control is one out of a multitude of useful used means in industrial check on. Due to the influence of factors such as the friction of the liquid itself, the control object has certain hysteresis and nonlinear characteristics. PID control is one of the most wide-ranging applied control algorithms and its wide operating range and simple function. In this paper, the modeling and simulation of representative three-tank PID and the traditional PID control and fuzzy PID control are compared and summarized.

The equipment procurement contract performance evaluation is an important means to effectively solve the problem like "inefficient progress", "low targets", and "increased costs" in equipment procurement. This paper defined the concept and connotation of the performance of equipment procurement contract, and proposed the equipment procurement contract performance evaluation, including influence factors analysis, evaluation indicators establishment, evaluation model construction, calculating results, etc., constructed four-dimension equipment procurement contract performance evaluation indicator system consisting of quality, schedule, funding, and service, and constructed the equipment procurement contract performance evaluation model based on BP neural network, and carried out a case analysis. The research conclusions provide a reference for the evaluation practices of equipment procurement contract performance.

By analyzing the basic theories and technical principles of managed pressure drilling, this paper establishes the cyclic pressure loss check and ECD calculation model under managed pressure conditions. The model is used to calculate and analyze the distribution of equivalent density curves in the wellbore under different drilling fluid densities and different pressure control conditions, and to analyze the distribution of ECD curves in pressure profiles at the same depth. Analyze the density range of the safe use of drilling fluid.

In order to reduce the cost, computer vision technology is introduced into the measurement of workpiece size and shape on the factory production line. At present, the most widely used solution is the neural network model based on big data. However, the lack of data and the high cost of data processing also greatly limit the practical application of this aspect. The method of feature extraction brings challenges to the real-time, rotation invariance, and anti-noise of online detection. In this paper, firstly, Harris operator is used to extract feature points quickly. Then a two-layer scale space based on causality is constructed to filter the noise and project downward to obtain the robust feature position, which provides a basis for subsequent processing.

Recommender systems are one of the most important technologies in the electronic commerce system. In a collaborative filtering recommendation algorithm, similarity calculation is the key to determining the efficiency of the recommendation algorithm. This paper analyzes the shortcomings of traditional similarity measurement methods in recommender systems and proposes a scoring-matrix-filling algorithm. Based on information categories and user interest similarity, the algorithm can reduce the negative influence of data sparsity on the recommendation result to some extent. The research results have certain practical and guiding significance.

With the emergence of data islands and the popular awareness of privacy, federated learning, as an emerging data sharing and exchange model, can realize multi-party collaboration under the premise of protecting data privacy and security because the data distributed in multiple devices cannot be sent locally. To achieve benefits for all parties involved, it has been widely used in many fields such as finance, medical care, and education. However, FL also has various security and privacy issues. Starting from the overview of federated learning, this article describes in detail the threat model and existing security issues, including replay attacks, poisoning attacks, reasoning attacks, etc., and then makes a certain analysis of FL privacy protection security technologies. Compared with SMC and HE, differential privacy is excellent in terms of efficiency. Finally, we discussed the challenges of privacy protection and security issues and future research directions.

The network car has solved some problems of traditional taxis from the perspective of safety and convenience, but there are still gaps in the inter-city field. This paper proposes a scheduling method to solve the inter-city travel field from the regional perspective. Firstly, the platform of the system is built from the perspective of software development, and then the traditional scheduling algorithm for dispatching orders in the city is studied. Finally, the intercity scheduling algorithm is mainly designed. This paper fills the gap in the field of inter-city travel between vehicles, and the scheduling algorithm performs well, with good order taking rate and travel efficiency. The establishment of the regional network car platform meets the needs of the public for inter-city travel. The platform itself has practical significance, and the study of scheduling algorithm also provides ideas and reference for the field of vehicle scheduling.

The application of virtual simulation technology promotes the reform of experimental teaching methods in colleges and universities to provide reliable support for cultivating students' professional engineering application ability and improve the overall quality of talent training. Based on real projects, this virtual simulation experiment project adopts the trinity teaching method of 'virtual environment, actual problems, real ability'. And it makes full use of natural language processing and other AI technology and software technology to provide students with requirement analysis experiments that are very close to reality. The practical environment has played a multifaceted effect and solved the students' neglect of the importance of requirement analysis.

Mobile edge computing(MEC) has already shown its powerful computing ability in mobile device which can broaden the capacity of computing. However, in view of mobile edge computing system, due to the varying network conditions and limitation of the wireless channel, the problem of long delay and high energy consumption maybe increase in the industrial production line. Therefore, computing offloading was proposed to reduce delay and energy consumption in mobile edge computing but the task allocation of offloading decision still can be improved by using different optimization algorithms. In this paper, we improved computation offloading method by an improved glowworm swarm optimization(GSO) algorithm to solve this problem for multi-user-multi-MEC in mobile edge computing. Compared with existing improved algorithms of computation offloading, the experimental results show that our proposal can reduce the system cost a lot(to 25%) which has a better performance of saving energy and reducing delay in the mobile edge computing environment.

Landslide is one of the most common disasters in nature and is a kind of slope failure. Many countries in the world suffered, because of its wide distribution, frequent occurrence and great harm. In China, the national economy and people's lives and property suffer heavy losses due to the landslide disaster. In this paper, first of all, we used the Bishop method to theoretically analyze the stability of a landslide in Sichuan Province, and then we established a simulation model in Flac3D, the rock and soil mechanics calculation software developed by ITASCA company in the United States, to analyze the stability of the landslide under complex conditions. Finally, the results of theoretical analysis and simulation are consistent, which can provide ideas and methods for geological disaster prevention in the future.

With the development of new technology, the traditional single application architecture has been unable to meet a variety of needs of the existing system, in order to solve the inconvenience monomer architecture system maintenance, a series of problems such as poor extensibility, deployment of trouble, micro service architecture is to break up the system, each service function is simple, can complete the function of single responsibility, and deploy alone, without centralized management, a service outage will not affect other service work, reduce the coupling between modules, the system performance improvement. Therefore, micro-service architecture is adopted to develop a new think tank evaluation system. In response to the needs of local think tank evaluation, this paper designs a new think tank evaluation system based on micro-service. The system adopts micro-service architecture to design and develop, and splits the services according to the business of the system. The system was divided into basic service, registration service, audit service, declaration service, review service, scoring service and log service. The communication between the services was realized through RPC remote invocation based on Dubbo distributed framework and Zookeeper distributed application. The system realizes the separation of front and rear ends, each service of the server is implemented with Spring Boot framework, the front end is implemented with vue.js framework, the system use Mysql database, and Redis is used to improve the system performance

IOP Publishing Limited ("IOPP") is retracting this paper following an investigation which has confirmed this work has significant overlap with a preprint written by another author group, without permission or acknowledgement to the original authors. IOP Publishing Limited request any citations to this article be redirected to the original work [1].

IOPP have made multiple attempts to contact the author but has not received a response. The author is encouraged to contact jpconf@ioppublishing.org.

[1] Wang Z, Malaney R and Burnett B 2020 https://arxiv.org/abs/2007.07748

Retraction published: 23 April 2021

This article has been retracted by IOP Publishing following an allegation that raises concerns this article may have been created, manipulated, and/or sold by a commercial entity. In addition, IOP Publishing has seen no evidence that reliable peer review was conducted on this article, despite the clear standards expected of and communicated to conference organisers.

The authors of the article have been given opportunity to present evidence that they were the original and genuine creators of the work, however at the time of publication of this notice, IOP Publishing has not received any response. IOP Publishing has analysed the article and agrees there are enough indicators to cause serious doubts over the legitimacy of the work and agree this article should be retracted. The authors are encouraged to contact IOP Publishing Limited if they have any comments on this retraction.

Retraction published: 9 September 2022