Research And Application Of Prediction Model Of User Churn Of Private Fund Wealth Management App

The research focus of this paper is to use the data in the internal database of a private fund wealth management app company in the current market as a research sample,analyze the user assets,user attributes and user behaviors,find the corresponding features that really affect the loss of users according to business experience,user characteristics and app use process,SPSS modeler was used for loss prediction modeling.Then,by evaluating the effectiveness of different data mining classifier algorithms,a model suitable for Internet private wealth management app users is selected as the final prediction model.Output the churn probability of each customer in the forecast period and circle the potential churn population.To help operators to intervene and retain potential lost customers in advance,so as to slow down the loss of users,increase the amount of positions,the number of positions and the income of the enterprise.The main contents and achievements of this paper are as follows:(l)This paper expounds and explains the research background and significance of this paper,summarizes the two methods of user churn research,and introduces the research of data mining for user churn prediction at home and abroad.And the related theories of the data mining algorithm used in the paper are sorted out and introduced.(2)Using the data in the company’s database,the classic data mining implementation process and theoretical knowledge are applied in the actual work.First,the specific process of selecting characteristic variables,data cleaning and establishing the data set required for modeling is described.Then smote algorithm is used to deal with the unbalanced data set in the training set.Finally,three commonly used data mining classification algorithm models are used for modeling.They are logistic regression model,decision tree model and random forest model.(3)By using the confusion matrix and its derivative indicators,ROC curve and AUC value,the effectiveness of the three user churn prediction models established in the previous article was evaluated.Use the data in the test set to verify the accuracy of the three types of algorithm models.After comparison,select a model with the best effect(decision tree model)and apply it to the actual work to verify its prediction effect on the user churn of private fund wealth management app.In practice,providing effective assistance to the users and operators of the existing private fund wealth management app is of great significance to the healthy development of the private fund wealth management platform.