Application of Decision Trees to Classify Loan Default Risk in Microfinance Institutions in Anambra State

Effective credit risk classification reduces non-performing loan ratios and improves the financial sustainability of microfinance institutions serving underserved communities in South East Nigeria. This study applied decision tree models to classify loan default risk among borrowers of three microfinance banks in Onitsha, Awka, and Nnewi, Anambra State, South East Nigeria. A dataset of 5,400 loan records was used, comprising borrower demographics, loan purpose, collateral type, repayment history, and business sector information. Class imbalance was addressed through stratified sampling. Classification trees using CART and C4.5 algorithms were compared against a logistic regression baseline. C4.5 achieved the best balance of precision (0.82) and recall (0.77) for identifying high-risk loans. The most discriminating split variables were previous default history, loan-to-income ratio above 0.6, and business sector. Agricultural loans showed the highest default rates at 28.4 percent, followed by petty trading at 24.1 percent. Collateral type was a significant predictor only for loans above N500,000. Rules extracted from the C4.5 tree were interpretable by loan officers without data science training. The study demonstrates that transparent, rule-based classifiers can be deployed in contexts where algorithmic accountability is important. Recommendations include deploying the decision tree as a loan assessment tool in microfinance branches, establishing a shared adverse credit registry for the Anambra microfinance sector, and incorporating seasonal agricultural risk adjustments into the default classification model.

Keywords: decision tree, loan default classification, microfinance, Anambra State, credit risk