Loan Risk Assessment for Umurenge SACCO using Machine Learning

Umurenge SACCOs are instrumental in fostering financial inclusion in Rwanda, yet they face significant challenges with high loan default rates that threaten their long-term sustainability. This study develops a predictive model using machine learning techniques to assess loan default risk among SACCO borrowers. Using a real, anonymized dataset of 2,000 loan applications from the Rwanda Cooperative Agency (RCA), we compare six machine learning algorithms: Logistic Regression, Decision Tree, Random Forest, Support Vector Machine, AdaBoost, and XGBoost. The study addresses class imbalance through balanced training approaches and evaluates models using accuracy, precision, recall, and F1-score metrics. XGBoost achieved the highest performance with 89.5% accuracy, while Logistic Regression demonstrated optimal balance between performance (86.5% accuracy, 85.2% F1-score) and interpretability, making it suitable for real-world deployment in SACCO environments. Key predictors identified include credit score, past loan repayment behavior, and monthly income. These findings provide a scalable, data-driven approach for SACCOs to transition from intuition-based to evidence-based credit risk assessment, supporting Rwanda's digital transformation goals in financial services.