An Analysis Of Customer Churn In The Telecom Industry Based On Data Mining Technology

In recent years, with the partition of the telecom industry in China, the top three telecom operators is busy at exploiting the market and occupying more market share. But they didn't pay enough attention to customer churn management about the existing customers. With the escalation of the customer churn rate, especially the loss of the high value customer, telecom industry appeared to be "more customers, less incomes". So, How to predict customer churn effectively and accurately, establish the strategy of customer retention scientifically and reasonably, and to great extents to reduce the customer churn rate, is a emergent key problem to telecom industry. It means a lot to increase customer churn rate, reduce operating costs, improve business performance, and finally realize the strategic shift from product-oriented to customer-oriented.The thesis's research goal is to combine the customer value theory of marketing and data mining technology and put into practice in the customer churn analysis in telecom industry. Based on the customer records in some China Unicom, the thesis uses the data mining software Clementine 12.0 which was owned by SPSS, takes the standard of CRISP-DM,and builds several classification models of customer churn prediction to predict the state of customers.For example,it builds C5.0, Neural Network,Bayes and so on. To continue,the thesis compares the models from numerical indexes to graphical indexes and forms a relatively optimal model.During the period of building the model,the thesis considers the feature of the unbalanced data in full,uses the method of data sampling and solves the problem effectively.Finally,the thesis improves the model by using the cost control technology based on classified by mistake from an economic cost standpoint and predicts the possible churn customers.Meanwhile, according to the characteristic of telecom industry, the thesis builds a feasible customer value evaluation index system. The thesis uses a improved method of AHP based on the experience of experts, gets the weight of every customer value evaluation index and finally has the relative customer value.Compared to the traditional customer churn analysis,it can judge the customer value more objectively,make the related departments take the customer rentention measures more effectively,and the cusomter retention more objectively. it finally reduces the cost of customer rentention.The meanings of thesis is not only lying in providing practical reference value to settle the customer churn problem of China Unicom, but also providing a beneficial reference to customer retention of telecom industry.