Method and system for predicting visitor losing

Abstract

The invention discloses a method and a system for predicting visitor losing. The method comprises: determining a target feature and an input feature, wherein the input feature is visitor behavior data, and the target feature is a visitor losing probability; dividing a data set for the visitor behavior data according to a preset ratio, wherein the data set includes a test set and a training set; calculating the penalty parameter lambda in the lasso algorithm for the training set by using a cross-validation method; training a lasso-logit model b using the training set according to the penalty parameter lambda; evaluating the lasso-logit model by using the test set; and finally outputting a prediction according to the evaluation result, so as to input the visitor behavior data into the prediction model, and output the visitor losing probability obtained by prediction. The invention is capable of effectively improving the effect of predicting visitor losing.

Description

technical field [0001] The invention relates to the technical field of communications, in particular to a visitor loss prediction method and a system applying the method. Background technique [0002] The advent of the information age makes the focus of enterprise marketing change from product center to customer center, and customer relationship management becomes the core issue of the enterprise. One of the key issues in customer relationship management is customer churn prediction. Based on customer behavior characteristic data, the customer churn behavior is predicted. The company formulates optimized personalized service plans for the lost and active users, adopts different marketing strategies, and improves User activity, enhance user monetization ability, and achieve the goal of maximizing corporate profits. [0003] A common analysis process in the aviation field is to use machine learning technology to predict the loss of online customers with the help of airline hi...

AI Technical Summary

Problems solved by technology

Usually, the determination of the algorithm needs to be combined with the actual business background. At the same time, if you choose an overly complex algorithm (such as a decision tree algorithm), it may lead to overfitting; if you choose an overly simple algorithm (such as a logit algorithm), it may lead to underfitting. Both overfitting and underfitting lead to a decline in the prediction effect of user churn

In addition, there is still a certain degree of subjectivity in which features are selected as the input of the predictive model.

Images