Neural network model: The churn score of each category email list user can be calculated, and the user operation can determine the operation strategy of users with different churn risks according to the score. Our modeling this time is to use the cox survival model for early warning, and we will write an article later to elaborate the application of the other two algorithm models. The model construction effect is as follows: What is the essence of lost user operations? Analyze how to build a user loss early warning system from three aspects Let's take a look at the effect of the model. The following is the actual lost customers of this sample analysis data. What is the essence of category email list lost user operations? Analyze how to build a user loss early warning system from three aspects.
The model case processing summary shows that 506 cases category email list are censored. This figure represents the number of customers who have not yet lost, accounting for 72.3%. (1) Which dimensions will affect the loss of users What is the essence of lost user operations? Analyze how to build a user loss early warning system from three aspects The final model calculates several impact indicators with strong correlation, including address, occupation, calling card service, wireless network service, wired network service, telephone time, etc. The Exp(B) value for an address means category email list that for customers residing at the same address, the annual churn risk is reduced by 100%−(100%×0.972)=2.8%.
Customers who have lived at the same address for category email list two years have a 100% lower churn risk − (100% × 0.9722) = 5.5%. The Exp(B) value of the calling card indicates that the risk ratio of churn for customers who do not subscribe to the calling card service is 2.024 times that for customers who subscribe to the service. The Exp(B) value for the web service indicates that customers who do not subscribe to the web service have a churn risk ratio of 0.577 times that of customers who subscribe to the service. (2) Survival curve of the average customer The Customer Survival Curve is a visualization of the model's predicted churn time for the "average" customer. The x-axis shows category email list when the event occurred.