Financial Fraud Transaction Prediction Approach Based on Global Enhanced GCN and Bidirectional LSTM

Money laundering is an act taken by criminals to cover up the nature and source of illegal gains. As money laundering data shows a complex time dependence, there is also a complex spatial correlation between different transactions. For this reason, we propose a financial fraud transaction prediction method based on global enhanced graph convolution and Bidirectional LSTM, called GEGCN-BiLSTM. First, BiLSTM is used to capture the time dependence in money laundering transactions. It not only considers the previous historical data, but also considers the information of subsequent time steps. Then, GEGCN is used to further mine the spatial global context relevance between different transactions. On each time stamp, the output information of GEGCN will be used as the input of BiLSTM to integrate time dependence and spatial dependence. The experimental results show that GEGCN-BiLSTM outperforms other comparison algorithms in terms of effectiveness and significance, providing a powerful tool for market transaction supervision.