A Transaction Fraud Detection Method Based on Heterogeneous Relational Network Attention Mechanism

Abstract

The invention relates to a transaction fraud detection method based on the attention mechanism of heterogeneous relational networks, the key of which is to extract entity relations according to transaction data, construct a bipartite graph of relational networks, and propose a method for homogenizing heterogeneous networks based on node shrinkage And the neighborhood information aggregation algorithm based on Attention mechanism. The credit transaction is modeled and analyzed from the perspective of the relationship network. At the same time, combined with the actual situation, the differentiated potential relationship between transactions is fully considered, and the Attention mechanism is designed to generate the differentiated relationship between nodes, which improves the effect of fraud detection. From the practical point of view, the method provided by the present invention conducts fraud detection by shifting the attention from the transaction node itself to the several-order neighborhood information of the transaction in the relationship network, and fully considering the potential difference between the transaction nodes.

Description

technical field [0001] The invention relates to a transaction fraud detection method based on a heterogeneous relation network attention mechanism, and belongs to the field of information technology. Background technique [0002] In recent years, with the development of Internet technology, online lending business has gradually become the main business model in the field of Internet finance due to its rapid review and low threshold for lending. However, due to the gradual acquisition of high-quality users by banks and other platforms, Internet lending companies continue to sink customers in order to expand their market share, resulting in increased credit risk, and the bad debt rate can generally reach about 30%. While online lending brings convenience to users, it also provides opportunities for online black production. In addition to assessing the creditworthiness of borrowers, it is also necessary to pay special attention to the occurrence of potential fraudulent activit...

AI Technical Summary

Problems solved by technology

[0010] Step 2. The credit transaction bipartite graph relationship network obtained in the previous step is a heterogeneous network. Aiming at the problem that the heterogeneous network cannot be represented uniformly, the node shrinkage algorithm is used to homogenize the credit transaction bipartite graph relationship network to obtain a homogeneous network H. Include the following steps:

Images