Broadband Customer Churn Analysis Based On Decision Trees Algorithm

With the growing of communication technologies in information age, competition in communication industry becomes more fiercer. The telephone's market are facing saturation while broadband access is widly available, both of which will lead competition to the Internet access market, and churn analysis about broadband customer will be a hot spot. So this paper tries to build a broadband customer churn model using the decision tree algorithm combined with Boosting technology. To inspection the model, a famous Cable Company's broadband customer churn is analyzed as an Example, then evaluation and suggestions are given to the model.This paper firstly introduces the research background, research question, research content and the structure of the paper. Then in the theoretical basis summary it introduces the related theories and means used in the research, which refers to CMS and decision tree algorithm. After a brief review on customer churn analysis's conception and existing technologies, solution of this paper is given—using decision tree C5.0 algorithm combined with Boosting technology and loss matrix to build a broadband customer churn alarm model. More details about the model's generating process and rulse are also illustrated. Then, based on the theoretic model and the customer churn historical data of Company S, a specific and practical model is generated according a commercial data mining procedure, followed by evaluation and suggestions on the model.The innovation of this study comes as follows:(1) customer churn analysis on a new application sub_field—broadband access service market; (2) this paper does a empirical analysis based on large amouts of real commercial data to test the theoretic model's practicality; (3) this paper uses boosting technology and loss matrix to improve the model's performance; (4) this paper respectively gives experimental and commercial evaluation on the model by using the indicators of Accuracy, Lift, Coverage Rate and Correct Rate; (5) this paper interprets the key factors affecting customer churn in broadband access service.The limitations of this study are as follows:(1)the choosing of model's input variables doesn't according to a scientific standard or procedure; (2) the searching of best parameters in the adjustment of model relies on manual trial and error; (3) it doesn't perform a complete and intensive customer churn analysis about Company S based on the gained model.