The Application Study of Credit Risk Model in Financial Institution via Machine-learning Algorithms
As a kind of statistical model, credit scoring technology is widely used in the risk assessment of loan applicants. It can predict the credit risk of applicants based on the information provided by the borrowers, such as their historical data and the data from banking system. Based on the data provided by a financial institution, this paper focuses on the credit risk analysis and evaluation based on machine learning algorithms, including Logistic Regression algorithm, Decision Tree algorithm, and Random Forest algorithm. What's more, this paper completes data preprocessing, variable selection, WOE coding discretization and credit scorecard creation. The basic score of this credit scorecard is 750 points. The probability of default will double when the score is reduced by every 60 points. The experimental results exhibit machine learning algorithms are feasible approaches to evaluate credit. Besides, within the study's training and testing samples, such objective evaluation parameters as precision, recall, AUC, KS, and F1 score, which indicate Random Forest algorithm, Decision Tree algorithm, and Logistic Regression algorithm can all apply to financial risk analysis.
Proceedings - 2020 7th International Conference on Information Science and Control Engineering, ICISCE 2020
First Page Number
Last Page Number
Wang, Yuanzhang; Lu, Jiongcheng; Qin, Jiehan; Zhang, Chenyi; and Chen, Yiyang, "The Application Study of Credit Risk Model in Financial Institution via Machine-learning Algorithms" (2020). Kean Publications. 1133.