Evaluation research of small and medium-sized enterprise informatization on big data

Under the background of big data, key construction of small and medium-sized enterprise informationization level was needed, but information construction cost was large, while information cost of inputs can bring benefit to small and medium-sized enterprises. This paper established small and medium-sized enterprise informatization evaluation system from hardware and software security level, information organization level, information technology application and the profit level, and information ability level. The rough set theory was used to brief indexes, and then carry out evaluation by support vector machine (SVM) model. At last, examples were used to verify the theory in order to prove the effectiveness of the method.


Introduction
Under the background of big data, small and medium-sized enterprise informatization level building needs to be focused on, but information construction cost will be high and whether the information cost inputs can bring benefit to small and medium-sized enterprises is still unknown. So how to effectively evaluate small and medium-sized enterprise informatization level is the research content in this paper [1]. SVM (Support Vector Machine, SVM) is a relatively new research results of artificial intelligence algorithms and has been used more and more widely. Its advantage is to overcome BP neural network fitting and easily into the local minimum problem when dealing with small sample data [2]. However, when SVM input data,the data redundancy cannot be determined, which increase the complexity of computing run. But rough set can effectively eliminate the noise in the input data [3]. So from this perspective, SVM algorithm and rough set theory are complementary inputting eliminating redundant indicators data to the SVM to train, and then predicting or evaluating test sample, which can reduce support in eliminating redundant information study and prediction burden, at the same time, improve the accuracy of the support vector [4]. Therefore, in this paper, the application of rough set and support vector machine (RS -SVM) are adopted to research on small and medium-sized enterprise informatization.

Model Building Description
In this paper, rough sets and support vector machine combined evaluation model is mainly used in the small and medium-sized enterprise informatization level evaluation, and the basic idea include application of theory of rough sets, eliminating redundant to effectively reduce noise of data, dividing the precise core attribute index data into the training sample and forecast sample and sending them to support vector machine (SVM) to learn and predict, and then evaluate.

Rough Set Theory and Attribute Reduction
Rough set theory is proposed by a polish mathematician Z. Pawlak in 1982, a new type of mathematical method dealing with vagueness and inaccuracy. When dealing with uncertain information, rough set does not need priority knowledge condition constraints, so it has been adopted in many fields.
In knowledge expression system It is not for the whole attribute set A, and only condition attribute set is reduced. Reduction of decision table is divided into two parts: The first step: reduction for attribute set: if C P  , P is independent of attribute set ,, and " The second step: reduction for attribute value: is consistent decision table, C P  is the reduction , of C. Value reduction targets at reduction P, or we can say that attribute value reduction is related with each decision rule on decision table. Any condition attribute value of rules in decision table can be reduced and after that it can be consistent with rules, that means no inconsistency occurs.

SVM theory
SVM is a new machine learning algorithm, on the basis of statistical learning theory, proposed by Vapnik in 1995. There is a data set: This data set can be distinguished by a hyper plane: , at the same time it should be the best hyper plane. A quadratic programming problem is introduced in the process of seeking for hyper plane and the equation is as below: (1) arrange optimize method was used to to solve the problems of the formula (1), and here is introduced into its dual problem, using nonlinear transformation(X) to tansfer original data set to the high dimensional space for data processing, and in this space for an optimal classification of the dual problem:

Rough set -support vector machine (SVM) model for small and medium-sized enterprise informatization level evaluation
Rough set can preprocess the index involved in small and medium-sized enterprise information system and eliminate unimportant attributes index from informatization horizontal correlation, which reduces the time complexity of support vector machine (SVM), accelerate Support vector machine (SVM) learning process. To evaluate the small and medium-sized enterprise informatization level through rough sets and support vector machine (SVM) model, first rough set makes attribute reduction for evaluation index system in order to reduce unnecessary workload for evaluation index weight, then using analytic hierarchy process and entropy value method to calculate index weight, and finally input comprehensive informatization evaluation scores into the support vector machine to forecast and evaluate. This paper first analyzes and researches on the influencing factors of small and medium-sized enterprise informationization level, collecting all kind of source data; next in the collected data, using rough sets to choose the data variables. At this stage, the obtained sample set removes all the unnecessary variables and only retain the variables impact prediction accuracy; Then, using the analytic hierarchy process and entropy value method to calculate index weight and get comprehensive informatization evaluation scores which transfer to the support vector machine prediction system to study and finally output the prediction results. After the model being built, in order to test whether it really can improve the prediction effect, this article uses the model to make empirical analysis.

Empirical Analysis
When filtering the evaluation index, this paper refers to the achievements of national informatization evaluation center and numerous research scholars, making comprehensive on the basis, and doing research on manufacturing enterprises from four aspects of informatization level of hardware and software security, information organization level, information technology application ability and profitability level.
This paper refers to enterprise informatization level evaluation index system of country small and medium-sized in reference [5][6] the system is as below table 1.  Table 1 shows that 40 specific indicators are included in this index system, such as information control ability level containing informationization plan ability level, informationization leadership level, information organization, information control, and information technology innovation ability and so on five indicators. Based on the original index system involved in table 1, we research the city's various types of 90 companies and obtain the firsthand data. Take information control ability level of indicators for example, as condition attributes, 5 sub indexes will obtain the original data processing, and according to the rough set theory, divide them into four different levels, {4,3,2,1 }, grade respectively mean for {excellent, good, acceptable and bad}. The evaluation criterion of the enterprise is also divided into four types, as decision attribute of rough set , =,3,2,1 {4}, grade respectively mean for {good, better, general, poor}. The rest 11 have the same processing way. This paper designed a kind of evaluation optimization strategy of rough set -support vector machines (SVM), overcoming the local minimum defects of the traditional BP network, and greatly improving the training speed. The enterprise informatization evaluation model was established based on rough setsupport vector machine (SVM). Evaluating by this model was more accurate and more convenient than BP neural network model, and ruled out the subjective randomness of selected factors to give weight value, making the evaluation results more objective, so under the background of big data, the research method in this paper has important significance in theory and practical application.