Graph classification-based arbitrage gang identification method

Abstract

The invention discloses a graph classification-based arbitrage gang identification method, which comprises the following steps of S1, obtaining data preprocessing, extracting entities and relationships of a composition, and constructing a knowledge graph; s2, performing group division on the constructed atlas by using a connected subgraph algorithm; s3, calculating risk index information of each group to form a business risk score; and S4, constructing and training a depth map convolutional neural network for predicting the structural risk score of the group. According to the method, a connected subgraph algorithm is adopted to carry out group division on the constructed atlas, so that the arbitrage gang can be found more easily; a depth map convolutional neural network is adopted to classify the whole map, and the risk of the whole group is directly predicted; according to the method, the structure risk of the group is judged through the depth map neural network, the business index is combined, the frequent set mining algorithm is used for analyzing the business index information of the group, the business risk score is calculated, and the interpretability of the model result is improved by combining the business data.

Description

technical field [0001] The invention relates to the technical field of electronic information, in particular to a method for identifying arbitrage gangs based on graph classification. Background technique [0002] With the rapid development of Internet technology, it has brought new opportunities and challenges to finance, e-commerce and other industries. Platforms and merchants can use the Internet to release various promotions online to attract users and increase traffic, but there are such a group of people who use various means to obtain discounts for marketing activities on multiple platforms, and have even formed a complete industry. chain, called wool party or arbitrage gang. These arbitrage gangs have caused huge losses to merchants and platforms. According to statistics, the economic losses caused by black industry arbitrage reach tens of billions every year. Existing methods for identifying arbitrage gangs are mainly based on expert rules and traditional machine ...

AI Technical Summary

Problems solved by technology

Existing methods for identifying arbitrage gangs are mainly based on expert rules and traditional machine learning models. Although expert rules are highly interpretable, they need to summarize and summarize historical risk events, and the summarized rules may vary from person to person. , poor consistency, slow response

The identification dimension of the traditional machine learning model is mainly a single user or a single merchant. It is difficult to identify such anomalies of a gang nature, because these arbitrage users are often no problem from the perspective of a single user, but users of an arbitrage gang are put together. unusual but obvious

Images