A Research Of Customer Churn Within Mobile Interne Background

With the development of mobile Internet applications,more and more traditional industries have achieved breakthrough development relying on it.The rapid expansion of the industry market and the homogenization have led to the customer churn.Based on this phenomenon,this article selects online travel industry data under the background of mobile Internet,through in-depth preprocessing and mining analysis of user behavior data,to achieve customer churn prediction,user profile analysis,and put forward relevant suggestions and measures.The problem of user churn studied in this paper is essentially a two-category problem.Through the study of the current general classification algorithm,this paper selects XGBOOST and Light GBM with good classification effect and strong explanatory ability as the classification model.Considering the impact of missing values,data imbalance,and variable redundancy on the final model results,this paper selects the Multiple Imputation By Chained Equations to fill in the missing values,uses the improved Borderline-SMOTE sampling method for imbalance processing,and selects a variety of features Engineering method for feature selection.The classic RFM model in the retail industry is selected as the basis of the user portrait model,and customer value,recent frequency,and consumption ability are selected as the main components of the characteristics to perform K-Means clustering to build a user portrait system.The research conclusion of this paper is that data preprocessing improves the final result greatly.User conversion rate and time factor have a greater impact on customer churn,and users are divided into five categories for analysis and marketing recommendations.The conclusion and relevant suggestions of this article will help the industry in the context of mobile Internet to strengthen customer relationship management,churn management and operation management in a targeted manner.The analysis and data processing ideas in this article can be used as reference for research on similar issues and have certain theories and practices significance.