Fraud Detection in Accounting and Finance Enhanced by Knowledge-Driven GAT Networks

Accounting fraud detection (FAFD) has recently emerged as a popular topic in academic circles, industry, and government agencies as a direct result of the correlation between the present economic situation and an increase in financial accounting fraud. Forensic accountants bring their specialized knowledge to the table to help unearth financial accounting crimes when internal auditing systems fail to detect red flags. This method entails three stages: feature selection, model training, and preprocessing. Business organizations’ financial statements, whether fraudulent or not, as well as letters to shareholders and financial news, are pre-processed using the CKIP system. In this proposed approach to use the simplest statistical technique during the features selection phase. Throughout training, it made use of a knowledge GAT. This innovative method outperforms Knowledge Graph and Attention Network, with an average accuracy of 95.43%.