A Comparison of Machine Learning Based Approaches in Predicting Agricultural Loan De-faulters Among Farmers in Lavun Local Government Area of Niger State

Abstract

Financial institutions in Nigeria have continuously extended generous loan offers to the manufacturing and industrial sector compared to the agricultural sector, due to the risk-benefit ratio difference attached to each. To aid the agricultural sector in Nigeria, the government established risk-sharing interventions in the agricultural sector with the aim of incentivizing the financial institutions towards issuance of credits to farmers. However, financial institutions still seek to reduce the leftover risk. This research was conducted in order to assist financial institutions reduce the risk of lending to farmers. A private agricultural loan dataset collected in Lavun Local Government Area in Niger state, Nigeria was used in this research to predict the likelihood of an agricultural loan default. Recursive feature elimination was used to reduce the features of the dataset from 60 to 44. Furthermore, machine learning algorithms of random forest, logistic regression, support vector machine, gradient boosting, and adaptive boosting were applied on the dataset. The results obtained shows that gradient boosting and random forest algorithms were the most effective in predicting agricultural loan defaults with precision and f1-score of 86.36% with 90.48% and 89.47% with 82.93% respectively. Improving the accuracy of the other machine learning models is proposed for further study.