Instance-based credit risk assessment for investment decisions in P2P lending

• We use an instance-based model to assess a loan’s credit risk.
• We formulate P2P lending into portfolio optimization with boundary constraints.
• We describe the similarity of two loans using default likelihood distance.
• We use kernel weighting to smooth risks of loans.

Recent years have witnessed increased attention on peer-to-peer (P2P) lending, which provides an alternative way of financing without the involvement of traditional financial institutions. A key challenge for personal investors in P2P lending marketplaces is the effective allocation of their money across different loans by accurately assessing the credit risk of each loan. Traditional rating-based assessment models cannot meet the needs of individual investors in P2P lending, since they do not provide an explicit mechanism for asset allocation. In this study, we propose a data-driven investment decision-making framework for this emerging market. We designed an instance-based credit risk assessment model, which has the ability of evaluating the return and risk of each individual loan. Moreover, we formulated the investment decision in P2P lending as a portfolio optimization problem with boundary constraints. To validate the proposed model, we performed extensive experiments on real-world datasets from two notable P2P lending marketplaces. Experimental results revealed that the proposed model can effectively improve investment performances compared with existing methods in P2P lending.