Skyler Speakman (IBM Research Africa): Three Population Covariate Shift for Mobile Phone-based Credit Scoring

Mobile money platforms are gaining traction across developing markets as a convenient way of sending and receiving money over mobile phones. Recent joint collaborations between banks and mobile-network operators leverage a customer's past mobile phone transactions in order to create a credit score for the individual. These scores allow access to low-value, short-term, un-collateralized loans. In this talk we will look at the problem of launching a mobile-phone based credit scoring system in a new market without either labeled examples of repayment or the marginal distribution of features of borrowers in the new market. The latter assumption rules out traditional transfer learning approaches such as a direct covariate shift. We apply a Three Population Covariate Shift method to account for the differences in the original and new markets. The three populations are: a) Original Market Members, b) Original Market Borrowers who self-selected into a loan product, and c) New Market Members. The goal of applying a generalized covariate shift to these three populations is to understand the repayment behavior of a fourth: d) New Market Borrowers who will self-select into a loan product when it becomes available.