A method and a device for predicting user loss of an insurance e-commerce platform

Abstract

A method for predicting user loss of an insurance e-commerce platform comprises the following steps: step 1, collecting original data of a website client, and storing the original data in a data warehouse according to a user rule; Step 2, performing cleaning, integration and protocol preprocessing on the original data, and further extracting variables for measuring the user value from the integrated data set for user subdivision; Step 3, selecting a reasonable observation variable and a proper user subdivision algorithm to carry out user classification; 4, selecting corresponding variables influencing user loss for different user groups, and performing user loss probability prediction respectively; 5, different prediction algorithms are selected for different types of users respectively, the model effect is evaluated through indexes such as the accuracy rate and the recall rate, and when the model effect is optimal, the final loss probability of the different types of users is output;And step 6, performing classified management on different types of lost user groups, performing group feature description respectively, and providing data reference for revocation strategy design so as to realize refined marketing and perform subsequent marketing effect analysis later.

Description

technical field [0001] The present invention relates to the field of user loss prediction, in particular to a method and device for user loss prediction of an insurance e-commerce platform. Background technique [0002] The trend of competition among e-commerce insurance websites is becoming more and more obvious, resulting in very similar services and low switching costs, so that users can easily jump between various insurance e-commerce providers, making the user status unstable and even leading to user loss. With the continuous refinement and deepening of marketing concepts, maintaining old users is particularly important for insurance websites. For this reason, companies need to effectively monitor and manage users, and predict user loss in a timely manner, so that various users with a tendency to lose , Take corresponding marketing measures early and in a targeted manner to retain old customers as much as possible. [0003] However, in the process of user churn predict...

AI Technical Summary

Problems solved by technology

[0003] However, in the process of user churn prediction, different types of users have different behavior habits, purchase preferences, self-worth, etc., especially in the special industry of insurance, the churn situation of different insurance types is different, and it is not possible to predict all user churn. "Equal treatment", for example, there is a significant difference in the pre-loss observation period between high-value customers and low-value customers. It is difficult to carry out targeted prediction work on insurance industry users in the existing loss prediction process. Therefore, the present invention is aimed at insurance e-commerce Platform user churn prediction, proposes a user churn prediction method based on user historical behavior and basic attributes and other data to classify users, realizes efficient, accurate and targeted user churn prediction, and continuously retains original users for enterprises, Minimize the probability of user churn

Images