Research On Kernel-based Credit Risk Model In P2P Lending

P2P (Peer to Peer) lending, i.e. individual personal loans, is a new kind ofnetwork-based lending. Investors face greater risks due to lacking of mortgage andcomplex loan relations in this kind of lending. As the data of P2P lending shows that,most of investors are at a loss. So risk assessment of P2P lending is of great practicalsignificance. Different from traditional lending, the relationship between investors andborrowers is many-to-many. Investors disperse risks by invest multiple loans and aborrower lends money from multiple investors. It is challenging to assess the risks andprofits of portfolio and optimize a portfolio in order to decrease the risk of P2P lending.There are few quantitative studies in this area.According to the characteristics of P2P lending, a quantitative definition of risks isgiven in this thesis. And based on data, a new Kernel model for credit risk assessmentincorporating mortgage default prediction is proposed. The assessment is performed attwo levels, namely, a single loan and portfolio. Investors can obtain quantitativeestimation for investments by using this model into the portfolio optimization. Moreover,the effectiveness of this model is validated by the real data in P2P lending.The new definition of P2P lending risks takes the potential loss of portfolio intoconsideration. The risk assessment model can precisely predict the risks and profits ofsingle loans and portfolios, which is the foundation of portfolio optimization. Afteroptimizing portfolios, quantitative estimation for investments can be obtained, i.e. gainprofits with the least risk by distribute the money into different loans.