Bankruptcy forecasting is crucial for financial stability and risk management. This study advances the accuracy of bankruptcy prediction using hybrid machine learning techniques on an imbalanced Polish dataset. The dataset presents challenges typical in real-world financial data, such as class imbalance and noisy features. Our approach combines the strengths of oversampling techniques, ensemble methods, and deep learning architectures to enhance predictive performance. Experimental results demonstrate significant improvements in precision, recall, and F1-score metrics compared to traditional methods. Insights gained from feature importance analysis provide deeper understanding of financial indicators driving bankruptcy. This research contributes to the field by proposing a robust framework adaptable to imbalanced datasets, offering practical insights for financial institutions to mitigate bankruptcy risks effectively.

Keywords: Bankruptcy forecasting, hybrid machine learning, imbalanced dataset, oversampling techniques, ensemble methods, deep learning, financial risk management