基于GBDT模型对于金融防欺诈的应用研究

Research on the Application of GBDT Model in Financial Fraud Prevention

摘要: 近年来，我国互联网消费金融飞速发展，促进消费需求，为低收入群体开辟借贷新途径，推动经济增长。但其快速扩张也导致借贷违约风险增加，风险特点异于传统金融，行业亟需创新违约风险管理、更新治理策略。本文用机器学习集成模型预测借款人违约风险：先研究互联网消费金融发展现状、借款人特征、违约情况及成因等；再分析消费金融欺诈分类、特点与防范，概述反欺诈模型；接着以Kaggle网站数据为基础，预处理后用XGBoost、LightGBM和CatBoost模型训练，获最优参数，建立综合评价模型实证分析；最后基于研究提出加强违约风险预测管理的建议。

Abstract: In recent years, China’s internet consumer finance has developed rapidly, boosting consumer demand, opening up new lending channels for low-income groups, and driving economic growth. However, its rapid expansion has also led to an increase in lending default risks, with risk characteristics differing from those of traditional finance. The industry is in urgent need of innovating default risk management and updating governance strategies. This paper uses machine learning ensemble models to predict borrowers’ default risks: first, it examines the development status of internet consumer finance, borrower characteristics, default situations and their causes; second, it analyzes the classification, characteristics and prevention of consumer financial fraud, and outlines anti-fraud models; third, based on data from Kaggle, after preprocessing, it trains XGBoost, LightGBM and CatBoost models to obtain optimal parameters and establishes a comprehensive evaluation model for empirical analysis; finally, it puts forward suggestions for strengthening default risk prediction and management based on the research.