Allocation of overdue loans in a sub-saharan Africa microfinance institution

Abstract

Microfinance is one strategy followed to provide opportunities to different economic classes of a country. With more loans, there is a high risk of increasing the loans entering the overdue stage, overloading the resources available to take action on the repayment. In this paper, it is only approached the experiment using clustering to the problem. This experiment was focus on a segmentation of the overdue loans in different groups, from where it would be possible to know what loans could be more or less priority. It showed good results, with a clear visualization of three clusters in the data, through Principal Component Analysis (PCA). To reinforce this good visualization, the final silhouette score was 0.194 which reflects that is a model that can be trusted. This way, an implementation of clustering loans into three groups, and a respective prioritization scale would be the best strategy to organize the loans in the team and to assign them in an optimal way to maximize the recovery.