An innovative model fusion algorithm to improve the recall rate of peer-to-peer lending default customers

• DOI: 10.1016/j.iswa.2023.200272
• Publication Date: 2023-08-25T09:48:46Z
• Authors: Zhida Liu; Zhenyu Zhang; Hongwei Yang; Guoqiang Wang; Zhenwei Xu
• Keywords: Default

Peer-to-peer (P2P) lending is a fintech innovation that provides loans to individuals and businesses without the need for additional intermediaries.However, if borrower fails repay loan on time, bank suffers financial loss due borrower's default. At present, many studies are trying improve accuracy of credit default risk prediction models reduce institutions' business, but it also meaningful study focus improving recall rate model results. In related research field prediction, crucial banks other institutions. The refers proportion all true positive examples correctly identified as examples. cases refer where borrowers default, while being means can accurately predict which may If banks' rates, this help them better assess risk, formulate appropriate policies, minimize losses. This aims further AUC metrics (area under ROC curve) P2P using dataset from Lending Club an improved machine learning fusion algorithm. Our proposed algorithm consists two algorithms. LightGBM XGBoost used obtain LGB-XGB-Stacking model. By optimizing evaluation in training phase these algorithms, we have achieved significant improvement results, especially defaulted customers overall metrics.After comparing predictive performance models, our following aspects. First, significantly than models. Second, outperforms metric. Among them, sample (default customer) 24.43% higher model, index 6.71% higher.In end, was found XGBoost, LightGBM, CatBoost performed very well terms improvement. rates three close, they Therefore, still effective method research, tree Although slightly lower above identifying defaulting customers. more identify bad debts