基于机器学习方法的企业贷款违约风险预测

Corporate Loan Default Risk Prediction Based on Machine Learning Method

摘要: 研究企业的贷款违约风险不仅对金融机构解决“惜贷”问题和防范信用风险具有重要的现实意义，而且能为企业规范自身经营和改善财务状况提出有针对性的建议及措施。本文根据某机构的企业贷款违约数据对贷款违约风险进行研究，首先对原始数据进行缺失值处理、特征选择和不平衡数据处理，然后利用逻辑回归、随机森林、XGBoost和LightGBM四种机器学习方法对数据进行建模和分析并比较模型优劣，最后利用GBDT模型计算特征重要性。结果表明：1) 三种集成模型的预测效果显著优于单一模型，2) 在集成模型中LightGBM模型表现出了最好的预测性能，3) 企业的纳税情况和曾经获得的授信情况可以作为判断该企业是否会发生贷款逾期现象的重要参考。

Abstract: The research on the loan default risk of enterprises not only has important practical significance for financial institutions to solve the problem of “reluctant to lend” and prevent credit risks, but also can put forward targeted suggestions and measures for enterprises to standardize their own opera-tion and improve their financial situation. This paper, based on the enterprise loan default data of an organization studies the default risk of the enterprise, first of all to the original data missing value processing, feature selection and unbalanced data processing, and then uses four machine learning methods of logistic regression, random forests, XGBoost and LightGBM for data modeling and analysis model, and advantages and disadvantage are compared. Finally, GBDT model is used to calculate the importance of features. The results show that: 1) The prediction effect of the three in-tegrated models is significantly better than that of the single model; 2) LightGBM model shows the best prediction performance among the integrated models; 3) The tax payment and the credit ob-tained by the enterprise can be used as an important reference to judge whether the enterprise will have the loan overdue phenomenon.