Small business loan risk assessment is inherently complex due to the interplay of quantitative and qualitative factors, data ambiguity and incomplete financial information. Conventional binary scoring models often fail to capture these nuances, leading to suboptimal lending decisions. To address these limitations, this study proposes a comprehensive fuzzy logic-based credit evaluation framework that integrates expert knowledge with data-driven optimization. The system incorporates eight key input variables: credit score, debt-to-income ratio, business age, annual revenue, cash flow adequacy, industry risk, employment growth and geographic risk. A Mamdani-type fuzzy inference system forms the core decision engine, while genetic algorithm optimization enhances membership parameters and rule weights. An initial expert-elicited rule base was expanded and refined using fuzzy c-means (FCM) clustering and multi-objective evolutionary optimization, resulting in a final set of 127 interpretable fuzzy rules. The framework was evaluated using 12,634 records from the Federal Reserve Small Business Credit Survey (SBCS) dataset. It achieved 84.7% classification accuracy and an AUC–ROC of 0.891, outperforming several conventional machine learning benchmarks while preserving intrinsic interpretability. Operationally, the system reduced projected credit losses by 25%, decreased false positives by 30.6% and improved processing efficiency by 81.9%. Fairness analysis demonstrated disproportionate impact ratios exceeding 95% across protected groups, indicating bias resilience. By combining predictive performance, transparency, and fairness, the proposed fuzzy–genetic framework supports equitable lending decisions, regulatory compliance, and informed managerial oversight, contributing to sustainable small business development and broader financial inclusion.
Nair et al. (2026) studied this question.