Forecast Of Customer Churn For China Mobile Communications Corporation

This paper establishes distance discriminant model,support vector machine,and neural network models for customer data of China Mobile Communications Corporation to predict and identify customers that may be lost,and to provide theoretical and methodological support for customer management and serial decision-making of communication companies.The main work is summarized as follows:1.The data are processed by missing values and abnormal values,and the clean data needed for analysis are obtained.2.Using K-means clustering,factor analysis based on random forest,logistic regression analysis and other methods,this paper explores the behavior characteristics of loss customers in China Mobile Communications Corporation.3.In this paper,the distance discriminant model is constructed with six independent variables: network length,monthly expenditure,individual degree,connection strength,individual information entropy and the change rate of individual degree.The discriminant accuracy is85.59%.4.Support vector machine model was constructed,and the discriminant accuracy was 85.46%.The accuracy of particle swarm optimization was87.55%.The accuracy of grid search optimization was 88.96%.5.A neural network model was constructed,and the discriminant accuracy was 85.95%,and 97.95% after optimization with AdaBoost algorithm.In this paper,the classification prediction model is applied to the identification of lost customers of China Mobile Company,and the model is constantly optimized in order to seek higher accuracy,reflecting greater practicability.In addition,we use various data analysis methods to explore the rules and characteristics of customer churn.Based on this,we propose retention strategies and marketing strategies.