基于过采样Logistic回归模型的互联网贷款违约预测研究  

Research on Online Loan Default Prediction Based on Oversampling Logistic Regression Model

作　　者：孙玮 周嘉莉 SUN Wei;ZHOU Jia-li(School of Finance,Hebei University of Economics and Business,Shijiazhuang Hebei 050061,China)

机构地区：[1]河北经贸大学金融学院,河北石家庄050000

出　　处：《华北理工大学学报（社会科学版）》2024年第1期54-61,共8页Journal of North China University of Science and Technology(Social Science Edition)

基　　金：河北省教育厅项目“教育数据资产交易机制研究”(SQ2021101);河北经贸大学项目“基于第三方量化平台的策略开发实验项目”(2021JYQ02)。

摘　　要：在持续增长的居民贷款消费需求刺激下,互联网贷款业务的规模呈现出持续快速扩张的发展态势,发挥机器学习模型在个贷违约预测的作用,控制和防范互联网贷款违约风险,具有十分重要的意义。通过对不同数据集的样本特征进行详细分析,构建个人信用风险评估指标体系,利用具有普适性特征和可解释性特征的Logistic回归模型对个贷违约进行预测。针对原始数据集存在不平衡样本的问题,分别采用过采样和欠采样的重抽样方法获得平衡样本集,调整正则化惩罚力度,选择最优结果的参数来进行建模,得到模型预测结果。最后对如何防范互联网贷款违约风险提出了相关建议。Under the continuous growth of consumer demand,the scale of online loan business is showing a trend of rapid expansion.It is of great significance to leverage machine learning models in predicting personal loan defaults,controlling and preventing the risk of online loan defaults.This article analyzes the sample characteristics of different datasets in detail,constructs a personal credit risk assessment system,and uses a Logistic regression model with interpretable features to predict personal loan defaults.To address the issue of imbalanced samples in the original dataset,oversampling and undersampling resampling methods were used to obtain a balanced sample set.The regularization penalty was adjusted,and the optimal parameters were selected for modeling to obtain the model prediction results.Finally,relevant suggestions were put forward on how to prevent the risk of default on online loans.

关 键 词：过采样 LOGISTIC回归模型 互联网贷款 违约预测 

分 类 号：F832.46[经济管理—金融学]