Predicting The Churn Status Of Internet Insurance Customers Based On Classification Models

In the last decade,as the Internet insurance business has entered a period of rapid development,the digital marketing system based on the Internet model is changing the way the insurance industry is driven by marketing,customer reach and connection,and even the development and value of insurance products and service solutions.Under the Internet insurance model,it is the frequency of use and excitement of customers in the insurance company’s private domain traffic(WeChat public number,small programs,APP,etc.)and whether they can meet the subsequent continuous marketing actions that are the new criteria for judging customer churn.Data mining based on big data technology is to extract useful information from the massive amount of customer data,usage behaviour,consumption behaviour,social information and other data to combine and correlate,to determine the Internet value of customers and the current status and tendency of churn,allowing companies to formulate customer service strategies in a timely and targeted manner,to create a precise marketing service body,to increase the activity of customers in the private domain traffic,thus further enhancing customer This will further enhance customer satisfaction and brand loyalty,and help insurance companies to create a new core competitive value and marketing value.Based on the relevant characteristics of Internet insurance business and customers,this paper constructs a model system combining an optimized customer value system model RFM and deep confidence network(DBN for short).From the new perspective of excited user churn and non-excited customer non-churn,the existing RFM model is optimised to identify different customer churn influencing factors;the DBN network is used to construct a churn prediction model for Internet insurance customers,and the DBN algorithm identifies different weights of specific parameters to find the churn factors of excited customers and the non-churn factors of non-excited customers,so as to compare and analyse the different value The relationship between the causes of churn for different types of Internet insurance customers is compared and analysed.The model was validated and empirically analysed using actual data,and compared with other prediction methods.The improved model is more efficient and can handle large amounts of data,and the accuracy and effectiveness of the model has been improved,allowing insurance companies to develop effective marketing strategies based on the relevant analysis results,and provide more valuable services to different types of Internet insurance customers.