Research Of Telecommunications Customer Segmetation System Based On Improved Clustering Algorithm

During the process of the telecommunications industry development,the homogeneous competition between major operators was more and more serious, which determined the competitiveness of enterprises also need to start from the differentiated operation, change to individual service, and finally implement on customer needs. Therefore, the customer segmentation was the top priority for the telecom enterprises to win, according to certain standards, the existing customers was divided into different customer groups with different characteristics, which was achieved by clustering analysis mostly in the actual market. K-means algorithm was a kind of classical algorithm in clustering analysis.Howerer, it needed to specify the number of target clusters in advance and was sensitive to the initial clustering center, which leaded to the poor stability in the existing telecom customer segmentation system. In view of the customer segmentation in the telecom industry, this paper had done the following work:Firstly, the concept and method of customer segmentation were introduced. This paper analyzed the selection scheme of segmentation variables and subdivision models with the characteristics of service. At the same time, it summarized and compared the clustering analysis. The classical K-means clustering algorithm was analyzed emphatically,followed with its implementation idea and the existing improvement scheme.Secondly, the functional requirements of each module of telecom customer segmentation system were given in the way of case figures, and the subdivision variables used in the system were described. The detailed design was made by the system hierarchy and the functions and data types involved in each implementation class. For the cluster analysis module, an improved DDK-means algorithm was proposed based on the traditional K-means clustering algorithm. The sample points with high degree of discretization in the high density region were chosen as the initial clustering centers, and the improved evaluation index CBWP was used to analyze the clustering results, by which we got the most accurate optimal clusters number.Finally, the improved DDK-mean algorithm was simulated and compared with the traditional K-means algorithm in the clustering accuracy, intra-class distance, with-class distance and stability. It was verified that the CBWP index performed well in getting the best number of clusters and the validity compared to other commonly used indicators.Implemented and displayed the telecommunications customer segmentation system through the actual data, and assessed the subdivision results and a sub-group characteristic.The DDK-means algorithm proposed in this paper was a further refinement of the original customer segmentation method. The customer segmentation system based on this improved algorithm could achieve more accurate customer segmentation and help to formulate more targeted marketing strategies, the most effective profit maximization, and better customer service.