Private Information, Credit Risk and Graph Structure in P2P Lending Networks

TL;DR

This paper explores how private communication and location data in P2P lending networks can be leveraged to improve credit scoring, with graph topology and location metrics significantly enhancing prediction accuracy.

Contribution

It introduces a novel approach using private network and location data, combined with graph analysis and machine learning, to better predict loan profitability in P2P lending.

Findings

Graph topology explains over 5.5% of profitability variability.

Location information accounts for an additional 19%.

Machine learning reduces mean squared error by 4%.

Abstract

This research investigated the potential for improving Peer-to-Peer (P2P) credit scoring by using "private information" about communications and travels of borrowers. We found that P2P borrowers' ego networks exhibit scale-free behavior driven by underlying preferential attachment mechanisms that connect borrowers in a fashion that can be used to predict loan profitability. The projection of these private networks onto networks of mobile phone communication and geographical locations from mobile phone GPS potentially give loan providers access to private information through graph and location metrics which we used to predict loan profitability. Graph topology was found to be an important predictor of loan profitability, explaining over 5.5% of variability. Networks of borrower location information explain an additional 19% of the profitability. Machine learning algorithms were applied to the data set previously analyzed to develop the predictive model and resulted in a 4% reduction in mean squared error.