Peer-to-peer (P2P) lending could facilitate more consumer access to credit. Due to the higher asymmetry of information, detection of charge-off/default risk for loans is challenging. The P2P industry could benefit from models that could more accurately predict loans that will result in default. The purpose of this study is to compare the performance of existing predictive models to a Heterogeneous Ensemble (HEE) model to determine if an HEE can improve the detection of loans that are likely to end in default. The study specifically compares Extreme Gradient Boosting, a neural network, and regression models against a HEE that combines the three models. The results indicate the HEE learning model provides a significantly higher detection rate of charge-off/default for loans than the three individual models.
Munmun, Mousumi and Booker, Queen, "Heterogeneous Ensemble Learning for Default Prediction in Peer-to-Peer Lending in USA" (2021). MWAIS 2021 Proceedings. 15.