2-stage modified random forest model for credit risk assessment of P2P network lending to “Three Rurals” borrowers

Abstract

With the rapid growth of the P2P online loan industry in the “Three Rurals” (agriculture, rural areas, and farmers) sector, it is imperative to manage the borrowing risk of borrowers in the rural areas. A credit risk assessment model is proposed to classify the credit worthiness of the “Three Rurals” borrowers. We select the loan data of the Pterosaur Loan platform as the research sample, and establish a 2-stage Syncretic Cost-sensitive Random Forest (SCSRF) model to evaluate the credit risk of the borrowers. From the random forest, we construct a cost relationship from the actual distribution of the data categories, introduce a weighted Mahalanobis distance using the entropy weight method in the cost function, and adopt a weighted voting for the cost-sensitive decision tree base classifier. The parameters of the SCSRF model are optimized via a grid search. We validate the SCSRF classification model against several established credit evaluation models.