Behavior prediction method and device based on artificial intelligence, terminal and storage medium

Abstract

The invention relates to the technical field of artificial intelligence, and provides a behavior prediction method and device based on artificial intelligence, a terminal and a storage medium, and themethod comprises the steps: clustering an original sample data set to acquire a plurality of group sample data sets; generating a first generation population according to the plurality of group sample data sets by using a first population generation model, with the original sample data in the same group sample data set corresponding to the same identifier; training the first generation populationby using an XGBoost model to obtain a plurality of predicted values, calculating head prediction accuracy according to the plurality of predicted values, and performing iterative optimization on theplurality of group sample data sets by using a genetic algorithm according to the head prediction accuracy until a target population is obtained; and constructing a sample data subset for the originalsample data of the first identifier in the target population, and training a behavior prediction model to predict the behavior of the target user. According to the invention, the target trained by the behavior prediction model can be consistent with the target of the business scene, and the accuracy of behavior prediction of the head crowd is improved.

Description

technical field [0001] The present invention relates to the technical field of artificial intelligence, in particular to an artificial intelligence-based behavior prediction method, device, terminal and storage medium. Background technique [0002] In the agent performance evaluation business, it is necessary to score the new prospective agents through the prediction model, and then take the low performance accuracy rate of the top 1% of the population with the highest score (easiest low performance) as the ultimate goal of the measurement model. Such people are not allowed to apply for joining the company, so the accuracy of low performance forecasts for these people is very high. [0003] In the process of realizing the present invention, the inventors found that most of the common prediction models based on machine learning use the log loss or cross entropy of the model as the loss function of model training, and the trained prediction model is based on the AUC index of t...

AI Technical Summary

Problems solved by technology

[0003] In the process of realizing the present invention, the inventors found that most of the common prediction models based on machine learning use the log loss or cross entropy of the model as the loss function of model training, and the trained prediction model is based on the AUC index of the overall sample sorting. However, in the agent performance prediction scenario, the AUC of the overall sample is not required to be too high, but there is a high requirement for the prediction accuracy of the head sample, which causes the model training Inconsistencies between the goals of the business scenario and the goals of the business scenario

Images