网络小额贷款业务个人信用风险评估——基于DNN-SMOTEENN-ExtraTrees组合模型  

Personal Credit Risk Assessment for Online Small Loan BusinessBased on DNN-SMOTEENN-ExtraTrees Combination Model

作　　者：吕秀梅[1] 张儒 LU Xiu-mei;ZHANG Ru(School of Finance,Chongqing Technology and Business University,Chongqing 400067,China)

机构地区：[1]重庆工商大学金融学院,重庆400067

出　　处：《数学的实践与认识》2023年第7期14-21,共8页Mathematics in Practice and Theory

基　　金：国家社会科学基金(19XYJ022)。

摘　　要：针对网络小额贷款业务,构建组合模型DNN-SMOTEENN-ExtraTrees评估网络小贷信用风险.首先利用SMOTEENN算法处理样本数据中“好”和“坏”样本分布极端不平衡情况,再利用极端随机数算法ExtraTrees对特征重要性进行评估并剔除无关变量,最后采用深度神经网络DNN评估网络小贷个人信用风险.通过召回率、精确度、F1值和AUC值等模型性能评价指标,与BP神经网络模型、Logistic回归及支持向量机比较,发现组合模型分类能力更显著,泛化能力更加优异,更适合数据规模大、维度高的网络小贷市场评估信用风险.Aiming at the small loan business of online company,this paper constructs a combination model DNN-SMOTEENN-ExtraTrees to assess the credit risk of online small loan.Firstly,SMOTEENN is used to deal with the imbalance of the distribution between“good"and“bad"samples in the sample data.Secondly,ExtraTrees algorithm is applied to evaluate the importance of features and eliminate the irrelevant ones.Finally,personal credit risk of online small loan can be assessed by DNN.Compared with BP neural network model,Logistic regression model and Support Vector Machine by index recall rate,accuracy,F1-Score and AUC-Score,it shows DNN-SMOTEENN-ExtraTrees model has better classification and generalization.Therefore,the model is more suitable for credit risk assessment of online small loan market.

关 键 词：信用风险评估 DNN-SMOTEENN-ExtraTrees组合模型 深度神经网络 SMOTEENN 极端随机树算法 

分 类 号：F724.6[经济管理—产业经济] F832.479F224