Research And Design Of Telecommunications Customer Segmetation System Based On Cluster Analysis

Along with the reform and developments of the telecommunications industry, the telecommunications industry's traditional monopolization has been broken, competition is becoming fiercer. Based on the principle of CRM(Customer Relationship Management), providing customers with end to end service process flow will be essential to the improvement of the telecom operator's crucial competitiveness, and will determine its prospect in this industry. Customer segmentation is the basis of implementing CRM principle and sophisticated operation. Making suitable and valuable customer segmentation, facilitating to provide every customer with specific services, aiding to create customer value and at the same time pursuing enterprise's profitable maximization have become key issues for telecommunications operators.As for customer segmentation in telecom industry, following work was accomplished.First, the current situation of the international and domestic customer segmentation was elaborated, and the features and shortcomings of the present mainstream data mining tools were described. According to the telecom industry's huge amount data and high multi-dimension, the principle of K-means was elaborated, the optimization of the initial value and the improvement of running efficiency were considered as the key points of the algorithm innovation. Therefore, the basic principle of k-m-d algorithm was presented and described.Second, the system framework of the business analysis system of China Unicom was described. Within this framework, according to the SEMMA(Sample-Explore-Modify-Model-Assess) from SAS, the customer segmentation system based on data warehouse was designed and implemented, which is running on the hardware platform of the business analysis as the subsystem of it, including data collection and preprocess module, clustering analysis module, model application module and system administration module.Finally, after collecting sample data, with some experimental parameters, the experiment was accomplished, the outcome of the experiment was analyzed in several aspect, it was proven that the innovated algorithm was efficient and effective,...