|
|
基于Bagging集成方法的互联网金融信用风险评估
|
Abstract:
互联网金融是对传统金融模式的延伸，但由于部分借款人在借款后无法按期、足额还款，使得互联网金融平台面临着信用风险。对借款人的信用风险进行准确评估，可以降低风险，并且能够在一定程度上为互联网金融行业的稳定发展提供保障。数据分析方法在信用风险评估领域已有广泛应用。本文从国内某互联网金融平台借款人的个人、资产、借款信息三类数据提取特征，研究了数据分析方法中Logistic回归的衍生方法逐步Logistic回归、弹性网络和Bagging集成方法的代表Bagging、极端随机树和随机森林。研究发现随机森林与逐步Logistic回归分别在F1-score、Accuracy、FPR和AUC指标下效果最优，且筛选出的重要特征也保持一致。
Internet finance is an extension of the traditional financial model. As some borrowers are unable to repay in full or on time, the platform faces credit risks. Accurate assessment of the credit risk can reduce losses and provide a guarantee for the stable development of the Internet finance industry. Data analysis methods have been widely used in the field. This paper extracted features from the three types of personal, asset and loan information. Stepwise Logistic regression and Elastic Net which are the derivative methods of Logistic regression, and the representative of Bagging ensemble method, Bagging, Extremely Randomized Trees and Random Forest were studied. It is found that Random Forest and Stepwise Logistic regression have the best results under F1-score, Accuracy, FPR and AUC indicators respectively, and the important features selected are also consistent.
| [1] | 粟麟, 杨伟明. 数字金融: 发展现状、未来趋势与监管启示[J]. 北方金融, 2021(6): 8-12. |
| [2] | 李焱文, 蒋文华, 王纯洁. 网络大数据信用风险评分能有效预测信贷违约风险吗? [J]. 经济问题, 2021(7): 70-77. |
| [3] | Moscatelli, M., Parlapiano, F., Narizzano, S., et al. (2020) Corporate Default Forecasting with Machine Learning. Expert Systems with Applications, 161, Article ID: 113567. https://doi.org/10.1016/j.eswa.2020.113567 |
| [4] | 方匡南, 章贵军, 张惠颖. 基于Lasso-Logistic模型的个人信用风险预警方法[J]. 数量经济技术经济研究, 2014, 31(2): 125-136. |
| [5] | 韦勇凤, 向一波. 基于Group-Lasso方法的非均衡数据信用评分模型[J]. 中国科学院大学学报, 2021, 38(2): 181-188. |
| [6] | 王小燕, 张中艳, 马双鸽. 基于文本先验信息的贷款信用风险评估模型[J]. 中国管理科学, 2021, 29(5): 34-44. |
| [7] | Xu, D.Y., Chen, J.H., Zhang, X.Y. and Hu, J.G. (2020) A Novel Ensemble Credit Scoring Model Based on Extreme Learning Machine and Generalized Fuzzy Soft Sets. Mathematical Problems in Engineering, 2020, Article ID: 7504764. https://doi.org/10.1155/2020/7504764 |
| [8] | Bekhet, H.A. and Eletter, S. (2014) Credit Risk Assessment Model for Jordanian Commercial Banks: Neural Scoring Approach. Review of Development Finance, 4, 20-28. https://doi.org/10.1016/j.rdf.2014.03.002 |
| [9] | 王程龙, 陈程. 基于决策树的P2P网贷平台信用评级体系研究[J]. 农村金融研究, 2016(12): 45-50. |
| [10] | Wang, G., Hao, J.X., Ma, J. and Jiang, H.B. (2010) A Comparative Assessment of Ensemble Learning for Credit Scoring. Expert Systems with Applications, 38, 223-230. https://doi.org/10.1016/j.eswa.2010.06.048 |
| [11] | Li, Y.H. and Chen, W.D. (2020) A Comparative Performance Assessment of Ensemble Learning for Credit Scoring. Mathematics, 8, 1-19. https://doi.org/10.3390/math8101756 |
| [12] | Tian, Z., et al. (2020) Credit Risk Assessment Based on Gradient Boosting Decision Tree. Procedia Computer Science, 174, 150-160. https://doi.org/10.1016/j.procs.2020.06.070 |
| [13] | 莫赞, 张灿凤, 魏伟, 游德创, 张舒. 基于Bagging集成的个人信用风险评估方法研究[J]. 系统工程, 2019, 37(1): 143-151. |
| [14] | 白鹏飞, 安琪, Nicolaas Fransde Rooij, 李楠, 周国富. 基于多模型融合的互联网信贷个人信用评估方法[J]. 华南师范大学学报(自然科学版), 2017, 49(6): 119-123. |
| [15] | 操玮, 李灿, 贺婷婷, 朱卫东. 基于集成学习的中国P2P网络借贷信用风险预警模型的对比研究[J]. 数据分析与知识发现, 2018, 2(10): 65-76. |
| [16] | 周永圣, 崔佳丽, 周琳云, 孙红霞, 刘淑芹. 基于改进的随机森林模型的个人信用风险评估研究[J]. 征信, 2020, 38(1): 28-32. |
| [17] | 廖理, 吉霖, 张伟强. 借贷市场能准确识别学历的价值吗?——来自P2P平台的经验证据[J]. 金融研究, 2015(3): 146-159. |
