| Current digital music industry in the "high speed" in the early stage of development,and the problems have gradually emerged,which is reflected in the increasingly common phenomenon of user churn and the gradual expansion of the scale of user churn,which has brought a certain negative impact on the normal operation of the digital music platform,and has been widely concerned by the platform management.How to reduce the use while maintaining steady development The loss of users,the improvement of user retention rate and the guarantee of user quality have become one of the most prominent and urgent problems that platform decision makers are facing.Therefore,this paper studies the problem of user churn of digital music platform.On the one hand,it enriches the research content of domestic digital music industry to a certain extent,on the other hand,it can provide support for the whole industry to formulate targeted policy recommendations.This paper reviews the research literature and methods of user churn at home and abroad,and finds that the algorithms suitable for user churn prediction in machine learning classification algorithms are random forest algorithm and xgboost algorithm,which are used to build a user churn prediction model,evaluate and compare the fitting effect of the models,and obtain the main factors affecting user churn.Based on the business practice,this paper designs the user value evaluation system from the current value and potential value of users,clusters the lost users and non lost users respectively,summarizes the user characteristics and user value,and gives targeted maintenance suggestions.The conclusions of this study are as follows:First,the random forest algorithm model and xgboost algorithm model after parameter optimization have better fitting effect on user loss prediction,but the prediction model based on xgboost algorithm is better than the performance of random forest algorithm on the whole,and the final accuracy of the model is 0.884.Second,the main factors that affect the loss of users include whether to renew the membership service automatically,whether to cancel the membership service actively,the number of user registration days,the number of user transactions,the fees and days paid in the actual transaction and the number of songs played by users.Third,based on the user value evaluation system,all users on the platform are grouped into six categories,among which,three categories of lost users are active lost users with high comprehensive value,demand lost users with medium comprehensive value and random lost users with low comprehensive value;three categories of non lost users are loyal old users with high comprehensive value and service users with medium comprehensive value and long-term users with low comprehensive value.Then,put forward corresponding user management suggestions through comparative analysis.In conclusion,through the empirical analysis of digital music platform users,it is of great theoretical significance and value to optimize the churn of the whole industry and understand the user behavior of Internet plus application in the new era. |
