Credit Risk Management in Microfinance: Application of Non-repayment Prediction Models

In microfinance, which plays an essential role in the economic and social development of every country, several risks threaten these institutions. In this article, we will study how to manage this type of risk using scoring through machine learning predictive analysis. To this end, we have conducted a detailed study of three predictive analysis methods (SVM; RF; KNN) and the tools needed to develop these algorithms using the Python programming language. We presented a descriptive and shape analysis to a database of 1001 companies applying for credit. This analysis enabled us to cleanse our dataset of all unnecessary data that could cause problems when running the classification algorithms using the Python simulation language. We concluded this article with an application of the three classification algorithms on the test data of our Dataset. This was followed by an evaluation of the accuracy scores, highlighting the Random Forest classifier with an accuracy score of over 98%.