Method and device for determining risk features, and account risk scoring method and device

Abstract

The application provides a method and a device for determining risk features, and an account risk scoring method and device. The method for determining risk features comprises steps of calculating the risk multiple of each value of features in an account sample; and calculating the fluctuation ratio of each risk multiple, and according to the fluctuation ratio, determining the features are risk features or not. The account risk scoring method comprises steps of determining risk features required in scoring by using the method for determining the risk features; and calculating the weight of evidence (WOE) of each value of the risk features, and according to the WOE, obtaining a risk scoring model. This application can predict the risk of the account from the perspective of data, and comprehensive evaluate the risk of the account according to the WOE of the value of each risk feature, so that the risk assessment of the account is more accurate and the dependence on human understanding in the risk assessment is reduced, and defects that the human understanding influences risk assessment results or the assessment is incomplete are tackled.

Description

technical field [0001] The application belongs to the field of account risk assessment, and particularly relates to a method and device for determining risk characteristics and an account risk scoring method and device. Background technique [0002] The risk judgment of an account is mainly based on the factual information in the history of the account, such as: CTU strategy gray list; trial center penalty list; cheating case storage list; other model outputs, such as highly confident batch registration, batch transfer , Batch login list; existing system blocking accounts, such as RDS captcha blocking accounts and so on. At present, there are mainly two types of account risk determination methods in the prior art: [0003] (1) Blacklist-based method: that is, the list of bad accounts obtained through the accumulation of lists in historical cases, including fraud cases, misappropriation cases, cheating cases, etc.; [0004] The blacklist method strongly relies on known list...

AI Technical Summary

Problems solved by technology

[0006] The rule-based method strongly relies on people's business understanding. With massive data, one-sided understanding is likely to occur. Without considering the dynamic changes of the business, some possible risks will be missed, resulting in inaccurate scoring results.

Images