Predicting credit risk in Peer-to-Peer lending with survival analysis

Abstract

Peer-to-peer lending is an online micro financing, where lenders and borrowers meet virtually for loan transactions. Along with the high growth, there is high risk in peer-to-peer lending due to uncollaterized loans, information asymmetry and lack of expertise on borrowers' creditworthiness. Therefore, it is highly necessary to analyze the credit risk of borrowers in peer-to-peer lending. Conventional credit scoring models can only classify loans into good and bad groups, but they fail to identify the timing of default. This paper proposes survival analysis approach to predict survival probabilities of loans in peer-to-peer lending at different time periods. The results from the survival analysis are effective in predicting the survival periods of the loans. Furthermore, the results from survival analysis are modeled for classification with neural networks.