Investment Decision Models In P2P Lending

P2P (Person-to-Person) lending allows individuals to lend to and borrow from each other directly on an Internet-based platform without the participation of traditional financial interme-diaries. As an emerging method of investment, P2P lending has created new challenges as to making effective investment decisions. To that end, in this dissertation, we develop three effec-tive investment models to enhance investment decisions in P2P lending.The first model is built from the investor composition perspective. Specifically, we inves-tigate whether investor composition could be used indicate the investment value and how to apply investor composition to improve investment decisions. To this end, we first build investor profiles based on a quantitative analysis of past performances, risk preferences, investment expe-riences and the reliability of investors. Then, based on investor profiles, we develop an investor composition analysis model, which can be used to select valuable investments and improve in-vestment decisions. To validate the proposed models, we perform extensive experiments on the real-world data from the P2P lending marketplace. Experimental results reveal that the investor composition model can effectively indicate investment value and help investors make better in-vestment decisions.The second model is developed from the borrower credit risk perspective. Specifically, we propose a kernel-based model that has the ability to assess each loan with both return and risk. In this model, we assess the expected return of each loan as the weighted mean of similar loans and the expected risk as the weighted standard variance of similar loans, where the weights are calculated by the mathematical framework of kernel regression. Using the kernel-based credit risk model, we formulate the investment decisions in P2P lending as a portfolio optimization problem with boundary constraints. To validate the proposed model, we perform extensive experiments on the real-world data from P2P lending marketplaces. Experimental results reveal that the proposed model can effectively improve investment performances over existing methods in the P2P lending market.Finally, we develop a multi-source loan evaluation model to effectively integrate multiple information sources for investment decisions in P2P lending. Specifically, we first we convert two information sources into two kernel weights and two correlation coefficients based on the mathematical framework of kernel regression and the importance of the information sources. Then, we integrate them into a unified multi-kernel weight to develop a multi-source loan e-valuation model for predicting both return and risk of each loan. Furthermore, based on the multi-source loan evaluation model and modern portfolio theory, we propose an investment al-gorithm for quantitative investment decisions. To validate the proposed model, we perform extensive experiments on real-world data. Experimental results reveal that the multi-source loan evaluation model can significantly enhancing investment performance over the existing model and other baseline models in the P2P market.