Early Warning And Value Analysis Of Customer Churn In 4S Stores

In today’s automotive industry,not only is there fierce competition in new car sales,but the after-sales market is also showing the same trend.For the after-sales business,the intensified competition has increased the risk of customer churn,and the customer churn rate has become one of the important pain points for car dealers.Therefore,accurate positioning of users who are prone to churn in advance allows dealers to adopt appropriate retention plans in a timely manner,thereby greatly reducing the corporate losses caused by the churn rate.This article organizes and summarizes the customer details of a 4S store from 2012 to 2019,and uses three machine learning models:decision tree,XGBoost,and LM neural network to predict whether customers are about to churn.By comparing the confusion matrix,F1 value,and AUC value of the three models,the LM neural network is comprehensively selected as the optimal prediction model,and the customer’s churn probability is calculated based on the optimal algorithm model.According to the different probability of churn,customers are divided into three groups:low-risk customers,medium-risk customers,and high-risk customers,so as to realize hierarchical early warning of customer churn.Finally,for customers at each risk level,an RFM customer value model is established,and personalized communication and effective recovery methods are developed according to different types of users,so as to provide companies with accurate and effective strategies.In the analysis process,this article mainly follows the steps of data descriptive analysis,feature engineering,model establishment,model evaluation,churn hierarchical warning,and user value segmentation.In particular,in the feature engineering part,the improved SMOTE and ENN comprehensive sampling method is mainly adopted to solve the problem of sample imbalance.Python has a wealth of data processing,data visualization and data modeling modules,which are powerful and can be directly called.In the analysis process of this paper,the modules of Pandas,Matplotlib,imblearn,Scikit-learn,keras,etc.in Python are mainly used for data processing and subsequent analysis.