Using Machine Learning Algorithms to create a Credit Scoring Model for mobile money users

Abstract

Statistical and artificial intelligence methods are used extensively to analyze credit and evaluate the credit risk of loan application clients. In this paper, mobile transaction data of agents from a digital payment switching company that is interested in offering microloans to its agents were used to determine the agent’s creditworthiness. Traditional credit scoring methods do not work for these agents as the transaction data is not recorded by a full-service financial institution like a bank. Different data manipulation techniques were explored to present data into features that can be used for scoring. The effects of resulting features were explored using correlation and singular value decomposition, and clustered using - means clustering to assess creditworthiness. After clustering agents into groups, these groups were clustered again to determine low-risk agents, and a formula to determine how much credit can be extended to low-risk agents was devised.