Analysis Of Customer Churn In E-commerce Based On Data Mining

The related survey results show that the number of Internet users and the growth rate of online shopping users in China has slowed down year by year.The “demographic dividend”that once brought great profits to e-commerce companies has gradually disappeared,and the cost of acquiring new users has continued to increase.Under such pressure,it is increasingly important for e-commerce companies to carry out customer churn management actively and improve customer retention.The key link for customer churn management is customer churn prediction analysis.Based on the above background,this essay carries out the analysis of customer churn for e-commerce by studying the relevant data of e-commerce enterprise customers.I hope that the research in this essay will more effectively improve the accuracy of customer churn prediction,maintain more customer resources,and improve the business economy.Due to the large scale of e-commerce customer data,this essay chooses the unique algorithm that has been popular in the field of data mining in recent years-XGBoost to build a customer churn prediction model.Aiming at the problem of misclassification that often occurs in binary classification problems,after referring to related literature,I introduce the “misclassification penalty coefficient” into the "loss function" to optimize the algorithm.The optimized algorithm effect is better than the previous prediction effect from any different evaluation index levels.Secondly,this article also considers the importance of customer segmentation to customer relationship management.The user's survival time "T" during the observation period is introduced to the traditional RMF customer value segmentation model so that a new value evaluation model RFMT model is constructed.Each category of customers is analyzed separately for customer churn to provide data support for the company's targeted customer relationship management for each type of customer,thereby implementing targeted customer relationship management strategies.Based on analysis of customer data of a domestic e-commerce company,the following results are obtained in this article:(1)the optimized XGBoost algorithm has improved AUC value and accuracy to 0.83% and 3.3%,(2)the customer churn prediction model has an excellent prediction effect for each type of customer after segmentation,(3)The optimized XGBoost algorithm has certain advantages over the churn prediction effect of C4.5,BP neural network,and SVM algorithm.