Cash Flow Dominance in SME Financial Distress Prediction: An Explainable AI Framework with 17-Feature Engineering and Sector-Level Risk Analysis for Indian Listed Companies

Indian Small and Medium Enterprises (SMEs) facechronic credit inaccessibility due to the absence of scalable,interpretable financial health assessment tools. Existing studies relyon small static datasets with manually selected ratios and binarydistress definitions that fail to capture gradations of financial risk.This paper presents a comprehensive three-class financial distressprediction framework applied to 135 BSE SME Exchange andNSE EMERGE listed companies across 7 sectors, spanning 1,350firm-year observations from 2015 to 2025. We engineer 17 financialindicators across five dimensions: profitability, leverage, efficiency,cash-flow quality, and growth. A composite distress score derivedfrom five data-driven warning conditions generates Healthy, AtRisk, and Distressed labels. Three classifiers — Logistic Regression,Random Forest, and XGBoost — are evaluated, with RandomForest achieving 92.96% accuracy and a weighted F1-score of0.930. SHAP (SHapley Additive exPlanations) analysis reveals thatcash-flow quality indicators — CFO/PAT, CFO/Debt, CFO/Assets,and CFO/Sales — collectively account for approximately 39%of model importance, establishing cash flow generation as thedominant predictor of financial distress in Indian SMEs. Sectoranalysis reveals Finance (27% distress rate) and Infrastructure(21%) as highest-risk segments, while Chemicals (0.5%) andEngineering (1.5%) demonstrate strong financial health. Thesefindings carry direct implications for fintech credit assessmentplatforms and SME lending policy in India.Index Terms—Financial Distress, SME, BSE SME Exchange,NSE EMERGE, Random Forest, XGBoost, SHAP, Cash Flow,Feature Engineering, Three-Class Classification, Sector Analysis,India, Explainable AI, Working Capital