Research Of The Mobile Customer Churn Prediction Based On The Data Mining Technology

In the era of big data, who can learn useful knowledge hidden behind the big data, and use up the tacit knowledge, who will have a chance to take the lead in finding business opportunities. Data mining is such a technology. It can extract people interested in from huge amounts of data. Appeared in recent years, many new methods of data mining, such as neural network, text mining, support vector machine (SVM), etc., especially in recent years, the basic concepts and methods of data mining has been forming, and gradually received the recognition from the people. The data mining research is to develop in the direction of deeper!With the continuous deepening of the reform of telecommunications, communications industry in China’s booming in recent years, the industry chain is becoming more and more complex structure, many links all affect the customer behavior, which is endowed with new connotation, customer churn to keep customer retention and customer. So, a lot of domestic telecom operators began looking for new ways, to predict the loss of telecom customer problems. Based on data mining technology in the telecom customer churn prediction research started in the domestic development together!Based on the theory of customer value and customer life cycle, with the aid of H city mobile company’s business data, using the Logistic regression and decision tree algorithm of data mining, in accordance with the CRISP-DM standard rule of data modeling, step by step according to business understanding, data understanding, data preparation, model building, model assessment and deployment steps, to move customer churn prediction research problem, and the loss of the management for mobile client provides strategic policy.The full text is divided into six chapters.The first chapter is introduction. In this article the research background, research current situation, main research contents, research methods and innovations are described.The second chapter is overview of data mining. Connotation of data mining, data mining tools, mining algorithm and modeling methodology is summarized, especially introduced the decision tree algorithm for mining and Logistic Regression; In combing CRISP-DM methodology, on the basis of data mining in detail the whole process of the use of Logistic regression modeling.The third chapter mobile customer churn analysis related theory. First of all to define the concept of mobile customer churn, determine the object of this study; Then, introduces the theory related to customer churn, namely customer value theory and the theory of customer relationship life cycle; Finally proposed the influence of qualitative move the reasons of the loss of customers, and classification in this paper.The fourth chapter is mobile customer churn the empirical analysis. Use H, moving the company’s business data, using Logistic regression and decision tree mobile customer churn prediction model is established, according to business understanding, data preparation, data exploration, variable selection, model construction and selection, evaluation and analysis of the data mining process model results, to move customer churn prediction research problem. Among them in variable selection, and puts forward the using value of WOE and IV value variable screening new method, by establishing the model of three effect comparison and analysis, it can be seen to the Logistic regression model with WOE values of input variables discretization processing, and IV value can also be used to filter prediction variables. In addition, three model comparison and analysis, by looking at their figure, can draw two conclusions:first, the use value of WOE discretization variables, can improve the Logistic regression model effect; second, using Logistic regression than using decision tree building mobile customer churn prediction model of the effect is better.The fifth chapter mobile is customer churn management strategy. According to the theory of the third chapter of customer value and customer life cycle theory, the fourth chapter the empirical analysis of customer churn prediction model as a result, the innovation in the potential customers into three categories, namely the probationary period of low value customer, plateau low value customers and degradation period high value customer, and customer retention strategy and marketing strategy respectively.The sixth chapter summarizes the full text, and provides Suggestions for future research.The innovation of this article mainly has the following two points: First, the customer churn literature review of this study shows that most of the scholars in customer churn prediction analysis, just using a data mining algorithm or no empirical analysis, the lack of using multiple comparison analysis between the modeling data mining algorithm. This article uses the Logistic regression comparative study on the two kinds of data mining and decision tree algorithm mobile customer churn problem. Through model comparison and analysis, more can explain the promotion effect of the model..Second, in a lot of using data mining techniques to predict customer churn in the study, the data mining modeling of variable selection process is not thorough. This article in the research, based on the degree of differentiation, the variables of WOE and IV value, correlation method to filter the independent variable, make choice according to the input variable.