An Intelligent AI-Driven Decision-Making Framework Integrating Machine Learning, Deep Learning, Data Analytics, and Emerging Computing Technologies
Abstract
Artificial intelligence is increasingly used for financial decision-making and early identification of corporate default and financial distress. This study developed an AI-driven early-warning framework integrating machine learning, deep learning, advanced data analytics, explainable AI, and probability-based decision support. The analytical pipeline included preprocessing, redundancy screening, feature selection, repeated stratified cross-validation, model comparison, SHAP-based interpretation, and risk stratification. Logistic regression, support vector machine, Random Forest, XGBoost, LightGBM, CatBoost, and a multilayer perceptron were evaluated using multiple classification and discrimination metrics. Random Forest achieved the highest mean ROC-AUC and PR-AUC, whereas CatBoost produced the highest accuracy, precision, specificity, and Matthews correlation coefficient. Overall, ensemble-learning models outperformed conventional statistical and neural-network approaches. Explainability analysis identified profitability, liquidity, leverage, financing burden, and asset-structure indicators as important contributors to default risk. Predicted probabilities were further converted into low, moderate, high, and critical risk categories, with the observed default rate reaching 96% in the critical-risk group. The findings indicate that integrating ensemble learning, explainability, and probability-based decision support can provide a robust and interpretable foundation for intelligent corporate early-warning systems. Future studies should validate the framework using larger multi-source datasets and evaluate cloud, edge, and distributed deployment environments.