Learning to recommend via random walk with profile of loan and lender in P2P lending

P2P Lending recommender systems are embracing portraying schemes to obtain profiles of both loan and lender, and thus to overcome inherent limitations of general recommendation models. A successful recommendation method requires proper handling the interactions between loans and lenders. We argue that three fundamental problems need to be addressed: 1) how to fully utilize different properties of loan for establishing its profile, 2) how to adapt social and psychological factors for enhancing lender's profile, and 3) how to exploit the interactions between loan and lender. To the best of our knowledge, there lacks a unified framework that addresses these problems. In this work, we contribute a new solution named RRWP (Recommendation via Random Walk with Profile of loan and lender), for learning recommender systems for P2P Lending. We develop a hybrid graph random walk-based model to capture the complicated interactions between loans and lenders. In particular, the algorithm consists of three stages for better P2P lending recommendation. (1) Loan profile is built by utilizing attributes of both loan and borrower; (2) Lender profile is established via his social and psychological factors together with interactions between loan and lender; (3) A hybrid graph is constructed based on which random walk is performed to recommend for both loan and lender. Extensive experiments on real-world dataset demonstrate the effectiveness of RRWP. Further analysis reveals that profile modelling is consistent with the basic investment theory in finance.