User risk classification method, apparatus, medium and apparatus based on machine learning

Abstract

The invention discloses a user risk classification method, apparatus, medium based on machine learning, belonging to the technical field of artificial intelligence. The method comprises the followingsteps: firstly, static characteristic data of a user which is not combined with network behavior and dynamic characteristic data which is combined with network behavior are obtained; Then, inputting the above two kinds of feature data into a pre-trained stochastic forest model after the target feature data is removed and the user risk level is output. Secondly, inputting the feature data after removing the feature data from the above two feature data and the user risk grade output from the stochastic forest model into the pre-trained logistic regression model and then the user risk score is outputted. Finally, classifying users according to the user risk rating output by the stochastic forest model and the user risk rating output by the logistic regression model. The invention classifies the user risk by using the user static characteristic and the dynamic characteristic data through the machine learning model, and effectively improves the accuracy of the risk classification.

Description

technical field [0001] The present disclosure relates to the field of artificial intelligence, in particular, to a user risk classification method, device, medium and equipment based on machine learning. Background technique [0002] User risk refers to the possibility that the user may suffer losses due to external factors or internal factors caused by the user's own behavior in a certain environment and within a certain period of time; the classification of user risk is based on the user's operating environment Under the external factors and internal factors, user risks are evaluated according to the possibility of risk. [0003] The existing user scoring system is basically a scorecard model based on user attributes or black-and-white list classification. Such an evaluation method based on static attribute information that does not combine user network behavior is not conducive to evaluating the risk of users in platforms with a large number of users. It will cause one-s...

AI Technical Summary

Problems solved by technology

[0007] The purpose of the present disclosure is to provide a user risk classification method, device, medium and equipment based on machine learning, and then at least to a certain extent overcome the limitations and defects of related technologies caused by the lack of comprehensive user characteristic data and single classification model The problem of inaccurate classification results caused by

Images