Fraud analysis method and system based on relational graph learning, medium and equipment

Abstract

The invention provides a fraud analysis method and system based on relational graph learning, a medium and equipment. The fraud analysis method based on relational graph learning comprises the following steps of: constructing a heterogeneous relationship graph according to an information record of a user; performing feature extraction on the information record and the heterogeneous relationship graph to obtain a plurality of user information features, and identifying an information entry behavior of the user through the user information features; respectively inputting the plurality of user information features into a primary classifier for primary classification; inputting all primary classification results into a secondary classifier to carry out classification decision so as to output aclassification result about whether the user is a fraudulent user; wherein the second-level classifier is an integrated classifier determined after the first-level classifier is subjected to model fusion. According to the method, the primary classifiers are trained by using the multi-dimensional features, and model fusion is performed on the primary classifiers, so that the performance index of fraud recognition is comprehensively improved.

Description

technical field [0001] The present invention relates to the technical field of fraud analysis, relates to a fraud analysis method, in particular to a fraud analysis method, system, medium and equipment based on relationship graph learning. Background technique [0002] At present, the prevalence of Internet fraud has even caused major losses to some trading users. However, Internet platforms have not been able to come up with a systematic solution for identifying fraudsters during the transaction process. They only use the user's transaction behavior and personal information to judge the user's credibility. Currently the most popular e-commerce Internet platforms, such as foreign eBay and domestic Taobao, are using a reputation evaluation mechanism based on feedback accumulation. Since the user's credibility is calculated by scoring, and this scoring is based on a simple feedback evaluation mechanism, fraudsters can create multiple accounts and improve their credibility thr...

AI Technical Summary

Problems solved by technology

[0007] In view of the shortcomings of the prior art described above, the purpose of the present invention is to provide a fraud analysis method, system, medium, and device based on relationship graph learning, which is used to solve the problem that the prior art cannot effectively combine various relationships between users. The problem of fraud detection

Images