Study Of Customer Churn Prediction In E-commerce Based On Data Mining

With the gradual disappearance of the "demographic dividend" and "mobile dividend" of the Internet in China,the competition in the field of Internet e-commerce is becoming increasingly fierce.It is difficult to continue the way of attracting and maintaining customers through low-price marketing.The precision marketing mode is gradually recognized by mature e-commerce companies.It requires deeper understanding and judgment of the customer's life cycle and loyalty.When the development of Big Data technology is less mature,the customer churn is mostly judged by the last consumption time of the customer,this division has a certain lag,and once the customer has left and then to take retention measures,the effect is often unsatisfactory.With the development of Big Data storage,data mining and analysis technology,it is possible to predict the customer who is going to churn based on data mining algorithm.When come to the operation of e-commerce platform,it is important to construct a customer churn prediction model based on data mining.On the basis of consulting a large number of literatures studies and research on customer churn,the author has carried out specific research on the topic of customer churn in the field of e-commerce.First of all,it elaborates the significance and necessity of e-commerce research on customer churn.Secondly,it briefly introduces the related concepts of customer relationship management and data mining,and discusses the theoretical concepts and formula derivation of Logistic algorithm and XGBoost algorithm in detail.Thirdly,the author select more than 20 key indicators related to customer churn,orders,reviews,login and check-in data from an e-commerce company and construct a churn prediction model through the Logistic regression and XGBoost algorithm.The model is evaluated by the outputting obfuscation matrix and accuracy,recall rate and ROC curve.The accuracy of the final customer churn prediction model is 76%.The importance of each variable factor in predicting user churn is analyzed by XGBoost model.It is proposed that the user's research method should be combined with the data mining result and find out the specific reasons why the customer gave up the website,and different marketing strategies should be adopted for different types of users to improve customer experience and save the company's operating costs then improve the operating profit of the enterprise.At the end of the paper,the main conclusions of this study are summarized,and the prospects for further research are presented.