Start Date
10-12-2017 12:00 AM
Description
P2P marketplaces provide a huge amount of transactional data for micro loan marketing analysis. Prior work primarily studies factors that reflect listings’ quality or affect lenders’ decision in a collective level; whereas what discriminative characters that an individual investor possesses and how individuals’ investment behaviors change over time are less studied. To this end, this article conducts a study from the individual investor level, namely investment behavior profiling. In particular, we first design a uniform and information-comprehensive feature representation to profile an individual ’s investment behavior at each time slot, which includes various attributes from the perspectives of investor, borrower, listing, investor-borrower relationship, and exterior factors. Based on the profile representation, we employ the recurrent neural network (RNN) to model individual investors’ long and short term time-varying behavior characteristics. Evaluations on real-life P2P datasets verify the effectiveness of our RNN method.
Recommended Citation
HAN, Xiao; Wang, Leye; and Huang, Hailiang, "Deep Investment Behavior Profiling by Recurrent Neural Network in P2P Lending" (2017). ICIS 2017 Proceedings. 11.
https://aisel.aisnet.org/icis2017/Peer-to-Peer/Presentations/11
Deep Investment Behavior Profiling by Recurrent Neural Network in P2P Lending
P2P marketplaces provide a huge amount of transactional data for micro loan marketing analysis. Prior work primarily studies factors that reflect listings’ quality or affect lenders’ decision in a collective level; whereas what discriminative characters that an individual investor possesses and how individuals’ investment behaviors change over time are less studied. To this end, this article conducts a study from the individual investor level, namely investment behavior profiling. In particular, we first design a uniform and information-comprehensive feature representation to profile an individual ’s investment behavior at each time slot, which includes various attributes from the perspectives of investor, borrower, listing, investor-borrower relationship, and exterior factors. Based on the profile representation, we employ the recurrent neural network (RNN) to model individual investors’ long and short term time-varying behavior characteristics. Evaluations on real-life P2P datasets verify the effectiveness of our RNN method.