A machine learning-based financial health assessment and credit scoring tool for start-ups and SMEs

Small- and medium-sized enterprises (SMEs) and startups play a critical role in economic growth, yet they often face significant challenges in accessing finance due to heightened credit risk and limited financial transparency. This study develops a machine learning-based framework for assessing the financial health and creditworthiness of SMEs, leveraging a corporate credit rating dataset. To enhance predictive performance, it integrates robust data preprocessing, dimensionality reduction via principal component analysis, and oversampling techniques, including SMOTE, ADASYN, SMOTE-Tomek, and SMOTE-ENN. PCA preserves 95% of the variance, with profitability and efficiency ratios emerging as the dominant determinants of financial health. Selected supervised learning models are trained using cross-validation and hyperparameter optimization. Model performance is evaluated using various metrics, with particular emphasis on recall to minimize false negatives. Results indicate that resampling substantially improve minority-class detection, with ADASYN offering the largest recall gains, especially for logistic regression. The framework provides a scalable and data-driven decision-support tool for financial institutions and SMEs.