Risk Management in P2P Lending Markets

P2P lending has garnered considerable attention and utilization owing to its minimal entry barriers, reduced expenses, and enhanced efficiency compared to conventional financial institutions. However, alongside its popularity, P2P lending markets also encounter risks, with personal credit risk emerging as the most salient concern. This study aims to explore the determinants that influence borrower risk preferences through the examination of loan data obtained from the Prosper platform. Using a fixed-effects regression model, this study examines the relationship between key variables and risk preferences. Temporal factors, especially weekdays and weekends, are found to be important factors influencing investor risk preferences. Additionally, this study utilizes machine learning algorithms to develop a default risk prediction model in P2P lending. Through rigorous comparative analysis and experiments, the random forest model demonstrates robust predictive capabilities. Furthermore, a combined learning model utilizing voting and bagging techniques is constructed by integrating random forest, linear regression, and Xgboost models. This ensemble model provides auxiliary support for P2P lending platforms in recommending investable orders to investors. The findings of this study provide valuable insights into risk management within P2P lending markets, particularly in terms of borrower risk preferences and the utilization of machine learning algorithms for risk prediction. The knowledge acquired from examining loan data from the Prosper platform carries practical implications for P2P lending platforms and risk management practitioners seeking to enhance risk assessment and control strategies.