Predicting Likelihood for Loan Default Among Loan App Borrowers: A logit Classification Approach

Predicting the likelihood of missed loan repayments is essential for banks and credit providers to effectively manage lending risks and promote responsible financial services. This research utilises a binary classification approach, specifically logistic regression, to 
categorise borrowers based on their potential to fail in meeting repayment obligations. The predictive framework includes a range of economic and personal attributes such as age, employment type, income level, loan size, repayment period, the number of current loans, applicable interest rate, and whether the borrower resides in a rural or urban setting. Data were obtained through a combination of loan repayment histories and structured questionnaires, covering a total of 397 individuals. Analytical procedures were executed in R programming. Key predictors influencing the probability of repayment failure include age, loan size, tenure, the count of existing credit facilities, and interest rate. Conversely, individuals with higher earnings were less likely to fall behind on repayments. The predictive tool demonstrated strong classification ability, achieving an overall correctness rate of 87.9%, a precision rate of 99.4%, a sensitivity of 88.1%, and an ROC, AUC value of 0.8029, suggesting solid differentiation between reliable and risky borrowers. As a result, the study advocates for more rigorous borrower vetting, loan structuring aligned with income capacity, and consideration of advanced machine learning models for refining credit risk analysis.