企業破產不僅對公司本身帶來重大的損失，還可能對整個行業、供應鏈及經濟體系造成影響，為維護公司聲譽及保護股東和投資者的利益，有效的破產風險預測及評估，無論對公司內部經理人或外部金融監管機關皆相對重要，因此過去已有多數研究透過不同的變數建立具高準確性的公司破產預測模型，又或者透過不同的可解釋人工智慧（Explainable Artificial Intelligence, XAI）套用於模型來解釋預測結果。而目前公司破產預測領域的研究亦多採用人工經驗或專業建立模型，再加以XAI工具進行預測解釋而未能結合實務需求，故本研究以台灣經濟新報(TEJ)資料庫所提供1999~2009年間台灣上市及破產公司做為應變數，高達93種財務比率做為自變數，採用客觀的機器自動學習(Auto Machine Learning)工具TPOT來建立最佳模型和超參數外，亦根據財務專業人員的建議，分析多種可解釋工具如SHAP、LIME、Anchor和反事實分析如何結合模型，協助管理者制定決策。
本研究發現透過局部可解釋工具中的Counterfactual(反事實分析)搭配視覺化圖形的方式，能使公司高層及外部監管人員清楚瞭解當前面臨財務困境公司當前所需改善的財務比率，以利後續提供改進方向及建議，此外透過全局可解釋工具SHAP結合特徵選擇，能相較過往傳統的特徵篩選方法具有更佳的特徵縮減效果。

Bankruptcy significantly impacts companies and their industry supply chains. Predicting bankruptcy is crucial for managers and financial regulators to safeguard reputations and protect stakeholders. Previous studies have built accurate bankruptcy prediction models using various variables and Explainable Artificial Intelligence (XAI) techniques for interpretation.
However, these often rely heavily on human expertise without addressing practical needs. This study uses data from the Taiwan Economic Journal (TEJ) on Taiwanese listed and bankrupt companies from 1999 to 2009, with 93 financial ratios as independent variables. It employs the Auto Machine Learning tool TPOT to establish optimal models and hyperparameters. Recommendations from financial professionals guide the use of interpretable tools like SHAP, LIME, Anchor, and Counterfactual Analysis to aid managerial decisions. The study reveals that Counterfactual Analysis, combined with visualized graphics, helps managers and regulators understand and address key financial ratios needing improvement. Additionally, using SHAP for feature selection proves more effective than traditional methods.

中文文獻
Su, S. (2021). XAI|透過LIME解釋複雜難懂的模型. https://tinyurl.com/mfvzmvkf
谭泽人. (2022). [可解释机器学习] Anchor解析. https://zhuanlan.zhihu.com/p/460410471
英文文獻
Antwarg, L., Galed, C., Shimoni, N., Rokach, L., & Shapira, B. (2023). Shapley-based feature augmentation. Information fusion, 96, 92-102.
Arrieta, A. B., Díaz-Rodríguez, N., Del Ser, J., Bennetot, A., Tabik, S., Barbado, A., García, S., Gil-López, S., Molina, D., & Benjamins, R. (2020). Explainable Artificial Intelligence (XAI): Concepts, taxonomies, opportunities and challenges toward responsible AI. Information fusion, 58, 82-115.
Brughmans, D., Leyman, P., & Martens, D. (2023). Nice: an algorithm for nearest instance counterfactual explanations. Data mining and knowledge discovery, 1-39.
Caouette, J. B., Altman, E. I., Narayanan, P., & Altman, Z. (1968). 1.6 Z-score estimated for non-manufacturers & emerging markets. Altman Z-Score, 189(1).
Casalicchio, G., Molnar, C., & Bischl, B. (2019). Visualizing the feature importance for black box models. Machine Learning and Knowledge Discovery in Databases: European Conference, ECML PKDD 2018, Dublin, Ireland, September 10–14, 2018, Proceedings, Part I 18,
Chawla, N. V., Bowyer, K. W., Hall, L. O., & Kegelmeyer, W. P. (2002). SMOTE: synthetic minority over-sampling technique. Journal of artificial intelligence research, 16, 321-357.
Cho, S. H., & Shin, K.-s. (2023). Feature-weighted counterfactual-based explanation for bankruptcy prediction. Expert systems with applications, 216, 119390.
Elhoseny, M., Metawa, N., Sztano, G., & El-Hasnony, I. M. (2022). Deep learning-based model for financial distress prediction. Annals of Operations Research, 1-23.
Ellson, J., Gansner, E., Koutsofios, L., North, S. C., & Woodhull, G. (2002). Graphviz—open source graph drawing tools. Graph Drawing: 9th International Symposium, GD 2001 Vienna, Austria, September 23–26, 2001 Revised Papers 9,
Explainable Artificial Intelligence (XAI). Defense Advanced Research Projects Agency. Retrieved May 5th from
Galli, S. (2021). Feature-engine: A Python package for feature engineering for machine learning. Journal of Open Source Software, 6(65), 3642.
Hutter, F., Kotthoff, L., & Vanschoren, J. (2019). Automated machine learning: methods, systems, challenges. Springer Nature.
Joshi, S., Ramesh, R., & Tahsildar, S. (2018). A bankruptcy prediction model using random forest. 2018 second international conference on intelligent computing and control systems (ICICCS),
Keya, M. S., Akter, H., Rahman, M. A., Rahman, M. M., Emon, M. U., & Zulfiker, M. S. (2021). Comparison of different machine learning algorithms for detecting bankruptcy. 2021 6th International conference on inventive computation technologies (ICICT),
Kim, H., Cho, H., & Ryu, D. (2022). Corporate bankruptcy prediction using machine learning methodologies with a focus on sequential data. Computational Economics, 59(3), 1231-1249.
KUMAR, D. (2022). Credit Card Fraud Detection-Different Models. https://www.kaggle.com/code/dhirajkumar612/credit-card-fraud-detection-different-models
Lee, Y.-G., Oh, J.-Y., Kim, D., & Kim, G. (2023). Shap value-based feature importance analysis for short-term load forecasting. Journal of Electrical Engineering & Technology, 18(1), 579-588.
Li, W. G. (2014). Corporate financial distress and bankruptcy prediction in the North American construction industry. Duke University, Durham, North Carolina.
Lundberg, S. M., & Lee, S.-I. (2017). A unified approach to interpreting model predictions. Advances in neural information processing systems, 30.
Mannila, H., Toivonen, H., & Inkeri Verkamo, A. (1997). Discovery of frequent episodes in event sequences. Data mining and knowledge discovery, 1, 259-289.
Meulemeester, B., De Oliveira, R. M. B., & Martens, D. (2023). Calculating and Visualizing Counterfactual Feature Importance Values. arXiv preprint arXiv:2306.06506.
Molnar, C., Casalicchio, G., & Bischl, B. (2018). iml: An R package for interpretable machine learning. Journal of Open Source Software, 3(26), 786.
Mustaqim, A. Z., Adi, S., Pristyanto, Y., & Astuti, Y. (2021). The effect of recursive feature elimination with cross-validation (RFECV) feature selection algorithm toward classifier performance on credit card fraud detection. 2021 International conference on artificial intelligence and computer science technology (ICAICST),
Olson, R. S., & Moore, J. H. (2016). TPOT: A tree-based pipeline optimization tool for automating machine learning. Workshop on automatic machine learning,
Olson, R. S. a. B., Nathan and Urbanowicz, Ryan J. and Moore, Jason H. (2016). Evaluation of a Tree-based Pipeline Optimization Tool for Automating Data Science. http://doi.acm.org/10.1145/2908812.2908918
Regenwetter, L., Abu Obaideh, Y., & Ahmed, F. (2023). Counterfactuals for Design: A Model-Agnostic Method for Design Recommendations. ASME 2023 International Design Engineering Technical Conferences and Computers and Information in Engineering Conference,
Ribeiro, M. T., Singh, S., & Guestrin, C. (2016). Model-agnostic interpretability of machine learning. arXiv preprint arXiv:1606.05386.
Ribeiro, M. T., Singh, S., & Guestrin, C. (2018). Anchors: High-precision model-agnostic explanations. Proceedings of the AAAI conference on artificial intelligence,
Shelke, M. S., Deshmukh, P. R., & Shandilya, V. K. (2017). A review on imbalanced data handling using undersampling and oversampling technique. Int. J. Recent Trends Eng. Res, 3(4), 444-449.
Shin, K.-S., Lee, T. S., & Kim, H.-j. (2005). An application of support vector machines in bankruptcy prediction model. Expert systems with applications, 28(1), 127-135.
Stańczyk, U. (2015). Feature evaluation by filter, wrapper, and embedded approaches. Feature selection for data and pattern recognition, 29-44.
Tsogoo, M. (2021). Bankruptcy prediction with explainable machine learning methods Masarykova univerzita, Ekonomicko-správní fakulta].