A Comparison of Different Rules on Loans Evaluation in Peer-to-Peer Lending by Gradient Boosting Models Under Moving Windows with Two Timestamps

• Published in: Computational economics Vol. 62; no. 4; pp. 1481 - 1504
• Main Authors: Zhou, Ligang, Ma, Chao
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.12.2023; Springer Nature B.V
• Subjects: Behavioral/Experimental Economics; Computer Appl. in Social and Behavioral Sciences; Credit scoring; Decision making; Decision support systems; Economic Theory/Quantitative Economics/Mathematical Methods; Economics; Economics and Finance; Interest rates; Investments; Loans; Math Applications in Computer Science; Net present value; Operations Research/Decision Theory; Peers; Present value; Rates of return; Peer-to-peer lending; Two timestamps; Investment rules; Gradient boosting
• ISSN: 0927-7099; 1572-9974
• DOI: 10.1007/s10614-022-10308-9

This study investigates different investment rules on loans in peer-to-peer lending, such as default probability (DP), credit grade from the platform, internal rate of return (IRR), net present value (NPV), and Sharpe ratio (SR). We use classical models and gradient boosting models (GBM) to evaluate loans in a practical setting by considering two timestamps associated with each observation in the data set collected from a P2P platform. The empirical study demonstrates the realistic performance of different investment rules. Furthermore, some investment decisions based on IRR, NPV, and SR can outperform those based on DP and credit grade from the platform, which may provide a P2P lending platform with an impetus for deploying decision support systems to help investors improve investment performance.