Cost-sensitive boosted tree for loan evaluation in peer-to-peer lending
作者：
Highlights：
• A novel cost-sensitive boosted tree model is proposed for loan evaluation in P2P lending.
• An improved portfolio allocation model is designed using integer linear programming with boundary constraints.
• Explanatory variables differ in predicting the probability of default and profitability for P2P loans.
• Traditional loan evaluation models may not guide profitable decisions.
• The proposed models outperforms several popular benchmark techniques in terms of profitability.
摘要
•A novel cost-sensitive boosted tree model is proposed for loan evaluation in P2P lending.•An improved portfolio allocation model is designed using integer linear programming with boundary constraints.•Explanatory variables differ in predicting the probability of default and profitability for P2P loans.•Traditional loan evaluation models may not guide profitable decisions.•The proposed models outperforms several popular benchmark techniques in terms of profitability.
论文关键词：Cost-sensitive learning,P2P lending,Loan evaluation,Extreme gradient boosting,Portfolio allocation
论文评审过程：Received 8 December 2016, Revised 20 June 2017, Accepted 20 June 2017, Available online 22 June 2017, Version of Record 29 June 2017.
论文官网地址：https://doi.org/10.1016/j.elerap.2017.06.004
