Telecom Industry Churn Warning

With the revolution of mobile phone technology,the telecom industry has more comprehensive and complex user behavior data.At the same time,with the development of big data field,the application of data analysis is imperative.At present,the most urgent application scenario is the fierce competition for the customer service competition,which is the most important work of early warning.How to determine whether the loss of users in advance,how to analyze the possible reasons for the loss of each type,as well as how to develop the relevant measures to retain,is now the most concerned operators.In this paper,the distributed computing platform Spark is used to build the analysis model.The data mining process as the main line of the statement,the design and development of two models.The first is to use random forest algorithm to generate telecom customer churn prediction model,model input and loss characteristics of the field,it can output the judge for the possible loss of the user list and its importance evaluation division of variable value;second is introduced into the prediction model has important characteristics of higher degree to K-Means clustering algorithm to construct loss the model of customer segmentation,customer churn model output accurate cluster segmentation results,through summarizing the properties and behavior characteristics of various cluster loss of users,they found the reasons for the loss of the corresponding,and formulate corresponding strategies to retain users.The application of the model is to match the loss of users by the cluster of the lost users,and the reasons for the loss of output.At the same time,the establishment of evaluation index system of customer value,and calculate the loss of pre user itself with the value of using the entropy method,operators can be put in the time and products according to market demand and their own profits reached a high value on pre loss of user retention,reduce economic losses and keep the cost of telecom operators,integrated into the customer segmentation matrix to provide a theoretical basis for the formulation of recommendations,and is expected to become a powerful operator control of market share.