Agricultural loan delinquency prediction using machine learning methods

The recent economic downturn in the agricultural sector that started 2013 has caused some concerns for farmers’ repayment capacity, which raises need precise prediction of financial stress sector. Machine learning been shown to improve predictions with large data, however, its application remains limited In this study, we approximate by loan delinquency, and predict it employing a logistic regression several machine methods. main datasets include Call Reports Summary Deposits from Federal Deposit Insurance Corporation (FDIC). Our results show ensemble methods have best performance accuracy, improvement 26 percentage points at most Naïve Bayes classifier is method maintain lowest cost false when failure identifying potentially high-risk loans very costly. From perspective banks, while there are important benefits using learning, bank-level costs also considerations may lead different choices