AI framework enhances bankruptcy prediction using ensemble models and XAI

Researchers have developed a novel framework for predicting bankruptcy by combining feature selection, hybrid resampling techniques, and stacking ensemble models with explainable AI (XAI). The study utilized the Taiwanese Bankruptcy Prediction dataset and found that different resampling strategies significantly impacted model performance, with SMOTE-ENN showing stronger detection of minority classes. Among standalone models, a GRU with SMOTE-ENN achieved the best predictive balance, while a hybrid stacking ensemble incorporating multiple machine learning classifiers and an LSTM model demonstrated the strongest compromise between sensitivity and specificity. Key predictors identified through SHAP analysis included leverage, profitability, solvency, and operational efficiency indicators, suggesting potential for more reliable early warning systems for financially distressed firms. AI

IMPACT Enhances early warning systems for financial distress by improving the accuracy and interpretability of bankruptcy prediction models.

RANK_REASON Academic paper detailing a novel methodology for bankruptcy prediction using AI techniques. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• LSTM
• SHAP
• SMOTE-ENN
• Taiwanese Bankruptcy Prediction dataset
• UCI Machine Learning Repository

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →