Aug 2026· Systems· Vol 14, pp. 923· 0 citations· 37 references
TL;DR
Findings indicate that explainable AI can support corporate bankruptcy early warning when predictive benchmarking is combined with transparent and auditable attribution analysis for financial decision-making.
Abstract
Corporate bankruptcy early warning has often been treated as a binary classification task, yet financial distress is better understood as the outcome of interacting financial conditions that must be interpreted within practical risk-management contexts. To address this issue, this study proposes a systems-oriented explainable artificial intelligence framework for corporate bankruptcy early warning. The framework implements a Hierarchical LFRM-MACI architecture in which the Local Feature Refinement Module (LFRM) refines representations within profitability, solvency, liquidity, efficiency, and growth/shareholder performance subsystems, and cross-subsystem attention models their interactions. The framework is evaluated on the public UCI Taiwanese Bankruptcy Prediction dataset under balanced and moderately imbalanced training settings. For reporting clarity, the benchmark methods are classified into traditional machine learning methods and deep learning methods; traditional tabular learners are included in the formal single-split empirical comparison, while the proposed method’s contribution is positioned as a structured deep representation with an attribution workflow rather than as an overall superiority claim over traditional machine learning models. In the single 70%/30% validation split, the proposed model obtains ROC-AUC values of 0.9286 and 0.9269 under the 1:1.0 and 1:2.5 settings, respectively. In repeated 5-fold cross-validation with five repetitions, its ROC-AUC is 0.8940 [0.8777, 0.9103] under 1:1.0 and 0.9135 [0.9001, 0.9269] under 1:2.5. To examine the interpretability of the predictions, Permutation Feature Importance (PFI) and SHAP are applied to identify subsystem-level attribution patterns across major financial subsystems. The explanation results highlight influential predictors associated with leverage pressure, profitability and asset structure, liquidity, operating efficiency, and growth/shareholder performance. These findings indicate that explainable AI can support corporate bankruptcy early warning when predictive benchmarking is combined with transparent and auditable attribution analysis for financial decision-making.
An Integrated Data-Driven Loan Management Framework that comprises an Artificial Neural Network (ANN) for credit default prediction, Structured Query Language (SQL) to systematically extract and transform data, and Interactive Visual Analytics Dashboards to aid in providing transparency within the decision-making process is discussed.
Ponsak .S. Bande, Chimuzuoroke E. Ugwuja, Blessing .C. Uzo et al.· International Journal of Lat...· 0 citations
By empirically proving that high-performance algorithms can be mathematically blind to demographic biases, this framework directly advances SDG 10 (Reduced Inequalities) and provides the accountable, feature-level justifications required for secure and sustainable financial inclusion (SDG 8).
Htet Nge Nge Ko, Aung Htoo Khine, Shadab Kalhoro et al.· Journal of Risk and Financia...· 0 citations
The evaluation of accounting transactions is increasingly challenging due to the growing volume of financial records, severe class imbalance, and the limited transparency of existing audit support systems. Many current machine learning approaches emphasize prediction accuracy while providing insufficient interpretability and weak support for risk-oriented audit decisions. To address these issues, this paper proposes an Intelligent Accounting framework based on explainable machine learning for risk-oriented transaction outcome prediction. The proposed framework integrates accounting-driven feature engineering, supervised learning, SHAP based explainable artificial intelligence, and probability-based risk scoring into a unified decision-support pipeline. Logistic Regression is adopted as the core predictive model due to its robustness, interpretability, and model parsimony under highly imbalanced transaction data. Experimental results on accounting dataset consisting of 1,000 transaction records show that Logistic Regression achieved the highest PR-AUC of 0.9737 and ROC-AUC of 0.6458 compared with Random Forest and XGBoost. The risk scoring mechanism also ranked problematic transactions within the highest-risk group, supporting audit prioritization. In addition, graphical SHAP analysis provides qualitative insights by identifying Operating Expenses, log_Operating Expenses, transaction timing, Transaction Volume, Profit Margin, Revenue, Expenditure, Cash Flow, Gross Profit, and Accuracy Score as influential factors affecting transaction outcomes. These findings show that the proposed framework not only predicts transaction outcomes but also explains the accounting factors behind each decision. Overall, this study transforms conventional transaction classification into an interpretable, risk-oriented, and audit-driven intelligent accounting system for transparent financial decision support.
J. K. Siregar, Astari Dianty, Antonius Bimo Rentor et al.· International Seminar on Int...· 0 citations
It is argued that predictive accuracy and regulatory transparency are not competing objectives but complementary necessities for institutional survival in Nepal’s cooperative sector.
S. K. Sahani, Tsair-Fwu Lee, Digvijay Pandey et al.· Journal of Intelligent Decis...· 0 citations
Standard credit-risk scorecards rely on linear ratio thresholds that break down when feature interactions are nonlinear and observations carry temporal dependencies. Qualitative signals embedded in corporate disclosures—tone shifts, forward-looking hedges, and sector-specific terminology—remain largely ignored by numeric-only models, even though such signals often precede ratio deterioration. This paper introduces a tri-modal deep learning framework that jointly trains three complementary branches: a Convolutional Neural Network (CNN) for cross-sectional ratio-pattern detection, a Long Short-Term Memory (LSTM) network for multi-quarter trend modelling, and a Natural Language Processing (NLP) branch for disclosure-text encoding. Prior to deep-model training, LASSO regularisation removes collinear financial indicators and SMOTE oversampling corrects the severe class imbalance characteristic of distress datasets. A feature-concatenation fusion layer integrates all three branch outputs; the resulting vector feeds a sigmoid classifier that produces a calibrated distress probability. Benchmarked against five baselines on four financial datasets, the model reaches 94.8% accuracy and 91.3% minority-class recall, with a 4.1-point F1 advantage over the strongest single-modality competitor.
Paiinti Meenakshi, Muthaluru Bhuvaneshwari, Jalla Ganesh et al.· International Conference Com...· 0 citations
Timely bankruptcy of the corporations is a major issue that investors, financial institutions, and regulatory bodies need to know in order to reduce economic losses and enhance decisions on risk management. Nevertheless, bankruptcy forecasting is difficult because of extreme imbalance in classes and nonlinear correlation between financial data. This paper suggests a machine learning model of corporate bankruptcy prediction, which is explainable and statistically justified through advanced ensemble learning methods. A comparative study was conducted on a financial dataset based on Logistic Regression, Random Forest, XGBoost, LightGBM, Tuned LightGBM, and Stacking Ensemble models with $\mathbf{6, 8 1 9}$ firms and $\mathbf{9 5}$ attributes. In order to solve the problem of data imbalance, threshold optimization was used, and the optimal decision threshold was obtained (0.13). Accuracy, Precision, Recall, F1-score, ROC-AUC, PR- AUC, Matthews Correlation Coefficient, and Brier Score were used to measure model performance. The optimized LightGBM model had a better performance with the following parameters: F1-score of 0.5124, MCC of 0.4999, ROC-AUC of 0.9549 and a Brier Score of 0.0234, which showed high discrimination and good probability calibration. The explainability of the proposed framework with the help of SHAP and the statistical test developed by McNemar additionally confirmed the strength and interpretability of the proposed framework, which is why it can be applied to real-world financial risk assessment.
Kanchan, Meenu Gupta, Rakesh Kumar et al.· International Conference on...· 0 citations