关键词: 关键词: 信用卡交易, 欺诈检测, 图神经网络, 自注意力变化

Abstract: Abstract: Addressing the issue of existing methods' inability to accurately identify intricate and diverse patterns of organized fraud, this paper proposes a new practical credit card fraud detection method based on an elaborate transaction graph to solve the problem of fraud detecti??on without dealing with gang fraud patterns. This method first constructs the association transaction graph from the original transaction information of the user, then uses the graph Transformer neural network layer to mine the gang fraud characteristics directly from the transaction network without cumbersome feature engineering. Finally, the high-precision detection of fraudulent transactions is realized by combining the topological features and time-series features in the fraud detection network. Test results of the credit card anti-fraud experiment show that the method proposed in this paper outperforms the benchmark model of 7 comparisons in most evaluation indexes. Average-Precision (AP) improved by 22% over the best benchmark in transaction fraud detection tasks; The area under the ROC curve (AUC) indicator increased by an average of 16%. These results indicate that the method designed in this paper is effective in the detection of credit card fraud transactions, can protect the financial security of credit cardholders and merchants, and lays a theoretical foundation for regulators and financial institutions to conduct fraud detection of credit card transactions.

Key words: credit card transaction, fraud detection, graph neural network, transformer