The Analyze of Relationship between Revenue and Customer Payment Methods in Small Medium Enterprise Based on Clustering K-Means

Business capital and revenue are not only the decisive of the health of SMEs but also they must be balanced. In general, customers find their benefit from the flexible payment methods while on the other hand the SMEs should get their benefit too. So that, it needs to be studied whether it is necessary for SMEs to get their profit in accordance to this situation. One of the methods that suitable to be applied is by applying customer groupings based on revenue and payment namely the K-means clustering method since it can raise several groups that have not been known before. This information is useful for SMEs to be utilized based on their needs. Data in this study were gathered from customer attributes, number of transactions, and payment methods. The number of centroids was 3. The grouping results were stopped at the 5th iteration. The finding showed that the ratio value of the 4th iteration and the 5th iteration having the same ratio value, 0.07393. From the results of the iterations can be found; first, based on the customers’ number, the groups can be classified into three C1(18%), C2 (45%), C3 (36%). Second, based on the average number of transactions, post-paid payments was in the first rank (12.7 / week). From the results, it can be analyzed that this situation is burdensome for SMEs because the more the number of transactions, the more investment must be prepared for accounts receivable.

1. Introduction SME is an asset for Indonesia, and is proven to survive the economic crisis. [1]. Therefore, it is important about how to help SMEs to develop themselves and solve their problems. [2]. The customer is an asset for Small Medium Enterprise (SME) [3], [4], where as an effort is the company to provide convenience for customers. Payment flexibility is one of the components needed by the customer, which can be an attraction for customers towards the company. However, the flexibility of payment for customers can IOP Publishing doi: 10.1088/1742-6596/1908/1/012021 2 have consequences for the company by providing more capital. If the consequences are balanced by the amount of profit the consequences are balanced by the amount of profit the company receives, then this is not a problem or vice versa. Of course, it is necessary to do a study of how the relationship of the number of customer transactions with the customer payment system, therefore companies can make more informed decisions for the continuity and development of their business. SMEs can use their transaction database to mine and analyze these needs. [5]. Most companies have realized that customer databases are very important resources that can be used to analyze customer characteristics to form appropriate marketing strategies and to regulate them. [6]. Customer segmentation is one of the activities commonly used by utilizing these data. In general, customer segmentation is defined as the process of grouping customers by being divided into several groups based on the characteristics and behavior of their transactions. [7]. Many studies have been conducted on customer segmentation. each study has a different purpose. One of study used K-Means find important ranked stock. [8], and other study to find customer segmentation user CRM and K-Means. [9]. This study has analyzed based on RFM (Recency, Frequency, Monetary) to find the potential customer. [10].
One approach is to extract data using the clustering method in data mining. Data mining contributes greatly to extracting hidden knowledge and information contained in the data used by researchers. [11]. The process of data mining in SME data is done by clustering using the K-means algorithm for customer segmentation. [12]. In research [13] has combined clustering methods with assembled algorithms to segment customer.

CLUSTERING METHODOLOGY IN DATA MINING
Knowledge discovery in Database (KDD) is defined as potential, implicit and unknown information extraction from a set of data [14], [15]. Datamining is in the KDD framework which is part of the component. K-means algorithm is an algorithm for grouping data that only applies to numerical data, so that if the data is not numeric normalization must be done. [16]. K-Means evective to reduce the euclidean distance from each object. [17]. Step. [17].
Several studies have been conducted among others: This study used the clustering method to cluster customer segmentation with RFM model. [18]. Other research tried to combine clustering method with decision tree for segmentation of customer. [19] [20]. Development of clustering with multiple attributes has also been done in marketing segmentation based on greedy heuristic. [21]. Segmentation of customer to analys of marketing use K-Means has done in [22] research. Other research, develop the clustering of customer segmentation to analys customer intention. [23] [24].

Research Methodology
The purpose model of clustering of customer segmentation with K-Means can describe in the Figure 2. Determine the number of cluster by determine K value 2.
Determine the centroid randomly 3.
Find number of cluster K 4.
Specify the distance of data to the controid = ∑ − (1) dij = Distance of object between object i and j P = dimension of data Xik = The coordinates of object i in dimension k Xjk = The coordinates of object j in dimension k BCV (Between Cluster Variation) = eucledien distance from m to mj BCV = d(mi,mj) WCV (Within Cluster Variation) = the smallest distance between data and centroid. WCV = ∑(the smallest distance between data and centroid) 2

5.
Data group based on the smallest distance

Experiment and Result
The dataset is extracted from data from one of the SMEs engaged in the sale of electric pulses. The first step is to do a proprocessing dataset. The dataset (Table-1) is carried out by the normalization process by making the category of payment system data an integer value.  The distance between centroids is calculated and then the shortest distance between the centers is selected as a determinant of the cluster group of data. (Table-3).  The final step is to find the ratio that is the result of between BCV (Between Cluster Value) and WCV (Within Cluster Value) in Table-3. The BCV value is 177.9547, and the WCV value is 5551.102. Therefore, ratio of Table-3 is 0.032058.

Section 1.02
Iteration-2 In the iteration-2 until the iteration-n has a different way of determining the cluster than iteration-1. In this step, the cluster is determined by calculating the average of the groups formed in the previous iteration, in Table-4.  Table 4b describe the result of data distance. In this iteration-2, it has produced changes in cluster groups from the previous iteration. This situation continues to be repeated until there is no change in the cluster group formed.  The result of the iteration-4 ratio is the same as the iteration-5. It can be concluded that the iteration process has been completed and the cluster result was produced by the last iteration. The result of customer segmentation based on clustering method as follows in Table 6: It can be analysed how payment method give the impact for transaction average. This study has compared the average of transactions in each cluster which are related to the payment method, can be seen in Table 7 and Graph-1.  Figure 3 : Chart of Comparison of cluster in payment methods and average transaction The grouping results have produced 3 clusters with the highest average transaction C2 use post-date payment (12.714), followed by C1 use Auto-Transfer payment (7.357), and C3 use Non-Auto-transfer and Combination payment (3.321). From the results of the grouping it appears that the biggest contributor to SME transactions is the customer with the post-date payment system. This data can be used as an important finding for SMEs to make decisions in the development and evaluation of their companies. From these results it can be analyzed that this situation is quite burdensome to SMEs because the more the number of transactions they will spend on the accounts receivable. Therefore, the clustering method is suitable for extracting data from companies to explore opportunities that might be used by companies including small and medium companies for evaluation and development.

Conclusion and Future Research
The clustering approach has been able to explore SME data to analyze possibilities that can be used as judgment to make decisions in the development and evaluation of the company. From the results of the iterations can be found; first, based on the customers' number, the groups can be classified into three C1(18%) is auto-transfer payment, C2 (45%) is post-date payment, C3 (36%) is non-auto-transfer and combination payment. Three (3) clusters formed resulted in the finding that the most transaction contributors were customers with a post-date payment system. Based on the average number of transactions, post-date payments was in the first rank (12.7 / week). This finding can provide advice for SMEs that many transactions by customers should also be balanced with the availability of capital. The smallest number of transactions is the customer with an auto-transfer system means that customers have 8 a tendency to be less interested in the payment system at the beginning, in other hand, they rely more on capital from SMEs. This can be a concern for SME managers to react to it. Determination of the number of clusters in this study is still done manually. In future work, it is necessary to think about the validity of the right number of clusters so that the results of grouping will be more accurate.

Acknowledgment
Author would like to thanks to everyone that supports this research starting from University Teknikal Malaysia Melaka (UTeM), Malaysia, DRPM DIKTI, Department Science and Technology (DST) under AIRFT India (RTF/2018/000033), Widyagama University of Malang, Indonesia, AR-PulsaBiz Reload Malang, Indonesia which is the one of SMEs in Indonesia, and for collegiate who gave some suggestions and valuable critiques for improving this paper.