Research on legal prevention and control of financial credit risk based on MoCo and GATv2 algorithm

This study proposes a novel framework for financial credit risk legal prevention and control, integrating MoCo (Momentum Contrastive Learning) and GATv2 (Graph Attention Network v2) algorithms to address challenges of complex interconnections and label sparsity. The framework first leverages MoCo for unsupervised pre-training to learn high-order semantic representations of nodes. Subsequently, GATv2's dynamic attention mechanism is introduced to precisely capture legal risk propagation paths. Evaluations on the public Lending Club and proprietary Fin-Law-CN datasets demonstrate superior performance. Specifically, on Fin-Law-CN, the proposed method achieved an approximate 4%-6% improvement in F1-Score compared to baseline models, with an AUC of 0.86, validating its effectiveness and robustness in identifying intricate financial legal risks.