Innovative Fraud Detection in Supply Chain Finance using Graph Attention Networks

Supply Chain Finance (SCF) is essential to streamlining supply chain capital activities and lowering management costs. Big Data, the IoT, AI, and Blockchain have helped SCF make better investment decisions and create personalized risk pricing models. Fast development of these technologies has led to a more nuanced distribution of unique financial fraud patterns among usual patterns and a huge increase of SCF data. Our method involves training, feature extraction, and model preparation. More data preparation was done after bias discovery and feature selection to prepare the data for machine learning models. Linear Discriminant Analysis (LDA) in feature extraction identified the descriptions' hidden subjects. Our models were trained using HGAT, or Hierarchical Graph Attention Network. Graph Neural Networks (GNN) and traditional HGAT were typically outperformed by the suggested model at 94.34% accuracy. This model can improve supply chain finance decision-making and risk management since our method detects subtle SCF financial fraud tendencies more accurately.