Consumer credit risk: Individual probability estimates using machine learning
作者：
Highlights：
• Default probabilities provide detailed information about consumer creditworthiness.
• The standard approach for estimating creditworthiness is logistic regression.
• We present a general framework to estimate credit risks using machine learning methods.
• We demonstrate probability estimation in Random Jungle, a fast random forests implementation.
• Random forests outperformed a tuned logistic regression on credit scoring data.
摘要
•Default probabilities provide detailed information about consumer creditworthiness.•The standard approach for estimating creditworthiness is logistic regression.•We present a general framework to estimate credit risks using machine learning methods.•We demonstrate probability estimation in Random Jungle, a fast random forests implementation.•Random forests outperformed a tuned logistic regression on credit scoring data.
论文关键词：Probability estimation,Random forest,Credit scoring,Probability machines,Logistic regression,Machine learning
论文评审过程：Available online 21 March 2013.
论文官网地址：https://doi.org/10.1016/j.eswa.2013.03.019
