BORSA İSTANBUL’DA FİNANSAL SIKINTI TAHMİNİ: MAKİNE ÖĞRENMESİ VE SHAP ANALİZİ

Abstract

Finansal sıkıntının önceden tahmin edilmesi; firma yöneticilerinin, yatırımcıların ve düzenleyici kurumların olası risklere karşı zamanında sağlıklı kararlar alabilmeleri açısından önem taşımaktadır. Bu çalışma, Borsa İstanbul’da işlem gören imalat firmalarının finansal sıkıntı durumunu bir yıl önceden tahmin etmeyi amaçlamaktadır. Bu kapsamda altı sınıflandırma algoritmasının performansı karşılaştırılmakta ve en yüksek tahmin performansını gösteren algoritma belirlenmektedir. Analizler, 251 firmaya ait 2.810 gözlemi içermekte; finansal sıkıntı, Zmijewski skoruna göre tanımlanmaktadır. Çalışmada lojistik regresyon, CART, rastgele orman, destek vektör makinesi, k-en yakın komşu ve yapay sinir ağı algoritmaları karşılaştırılmaktadır. Model geliştirme ve test dönemleri, zaman sırası korunarak oluşturulmakta; hiperparametre seçiminde genişleyen pencereli ileri-zaman çapraz doğrulama kullanılmaktadır. Sınıf dengesizliği nedeniyle temel seçim ölçütü olarak PR-AUC esas alınmakta ve seçilen algoritmanın tahminleri SHAP yöntemiyle açıklanmaktadır. Çalışmanın bulgularına göre rastgele orman, doğrulama sürecinde en yüksek ortalama PR-AUC değerini üretmektedir. Bu algoritma test döneminde 0,873 ROC-AUC, 0,385 PR-AUC ve 0,747 dengeli doğruluk değerine ulaşmakta ve finansal sıkıntılı gözlemlerin %62,5’ini doğru belirlemektedir. Eksik değer içeren gözlemlerin dışlandığı duyarlılık analizinde de aynı algoritma seçilmekte ve benzer performans değerleri elde edilmektedir. SHAP bulguları, tahminlerde özellikle sermaye yapısı, likidite ve karlılık göstergelerinin öne çıktığını göstermektedir. Bu sonuçlar, makine öğrenmesi yöntemlerinin finansal sıkıntının önceden tahmin edilmesine katkı sağlayabileceğini göstermektedir.

Keywords

• finansal sıkıntı tahmini
• makine öğrenmesi
• rastgele orman
• SHAP

FINANCIAL DISTRESS PREDICTION ON BORSA İSTANBUL: MACHINE LEARNING AND SHAP ANALYSIS

Abstract

Predicting financial distress in advance is important for enabling corporate managers, investors, and market regulators to make timely and informed decisions in response to potential risks. This study aims to predict the financial distress status of manufacturing firms listed on Borsa İstanbul one year in advance. To this end, the performances of six classification algorithms are compared, and the algorithm with the highest predictive performance is identified. The analysis includes 2,810 observations from 251 firms, and financial distress is defined according to the Zmijewski score. The algorithms compared are logistic regression, CART, random forest, support vector machine, k-nearest neighbors, and neural network. The model development and test periods are constructed by preserving the temporal order, while hyperparameters are selected using expanding-window time-series cross-validation. Given the class imbalance, PR-AUC is used as the primary model selection criterion, and the predictions of the selected algorithm are explained using SHAP. The findings show that random forest produces the highest average PR-AUC during validation. In the test period, this algorithm achieves a ROC-AUC of 0.873, a PR-AUC of 0.385, and a balanced accuracy of 0.747, correctly identifying 62.5% of financially distressed observations. These findings hold in a robustness analysis in which observations with missing values are excluded rather than imputed, where random forest is again selected and performance values remain similar. The SHAP results indicate that capital structure, liquidity, and profitability indicators are particularly prominent in the predictions. Overall, the results suggest that machine learning methods can contribute to the early prediction of financial distress.

Keywords

• Financial distress
• machine learning
• random forest
• SHAP

References

1. Aker, Y., & Karavardar, A. (2023). Using machine learning methods in financial distress prediction: Sample of small and medium sized enterprises operating in Turkey. Ege Academic Review, 23(2), 145-162. https://doi.org/10.21121/eab.1027084
2. Aksoy, B., & Boztosun, D. (2020). Comparison of machine learning methods in prediction of financial failure of businesses in the manufacturing industry: Evidence from Borsa İstanbul. Anadolu Üniversitesi Sosyal Bilimler Dergisi, 20(4), 237-268. https://doi.org/10.18037/ausbd.845792
3. Aktaş, R., Doğanay, M., & Yıldız, B. (2003). Mali başarısızlığın öngörülmesi: İstatistiksel yöntemler ve yapay sinir ağı karşılaştırılması. Ankara Üniversitesi SBF Dergisi, 58(04), 1-24. https://doi.org/10.1501/SBFder_0000001691
4. Alomari, Y., & Andó, M. (2024). SHAP-based insights for aerospace PHM: Temporal feature importance, dependencies, robustness, and interaction analysis. Results in Engineering, 21, 101834. https://doi.org/10.1016/j.rineng.2024.101834
5. Altman, E. I. (1968). Financial ratios, discriminant analysis and the prediction of corporate bankruptcy. The Journal of Finance, 23(4), 589-609. https://doi.org/10.2307/2978933
6. Andrade, G., & Kaplan, S. N. (1998). How costly is financial (not economic) distress? Evidence from highly leveraged transactions that became distressed. The Journal of Finance, 53(5), 1443-1493. https://doi.org/10.1111/0022-1082.00062
7. Aydin, N., Sahin, N., Deveci, M., & Pamucar, D. (2022). Prediction of financial distress of companies with artificial neural networks and decision trees models. Machine Learning with Applications, 10, 100432. https://doi.org/10.1016/j.mlwa.2022.100432
8. Barboza, F., Kimura, H., & Altman, E. (2017). Machine learning models and bankruptcy prediction. Expert Systems with Applications, 83, 405-417. https://doi.org/10.1016/j.eswa.2017.04.006

9. Beaver, W. H. (1966). Financial ratios as predictors of failure. Journal of Accounting Research, 4, 71-111. https://doi.org/10.2307/2490171
10. Bellovary, J. L., Giacomino, D. E., & Akers, M. D. (2007). A review of bankruptcy prediction studies: 1930 to present. Journal of Financial Education, 33, 1-42. https://www.jstor.org/stable/41948574
11. Cakiroglu, C., Aydın, Y., Bekdaş, G., & Geem, Z. W. (2023). Interpretable predictive modelling of basalt fiber reinforced concrete splitting tensile strength using ensemble machine learning methods and SHAP approach. Materials, 16(13), 4578. https://doi.org/10.3390/ma16134578
12. Campbell, J. Y., Hilscher, J., & Szilagyi, J. (2008). In search of distress risk. The Journal of Finance, 63(6), 2899-2939. https://doi.org/10.1111/j.1540-6261.2008.01416.x
13. Carton, R. B., & Hofer, C. W. (2006). Measuring organizational performance. Edward Elgar Publishing.
14. Ceyhan, İ. F. (2023). İşletmelerde finansal başarısızlık öngörüsünde makine öğrenmesinin kullanımı: BİST uygulaması. Finans Ekonomi ve Sosyal Araştırmalar Dergisi, 8(3), 564-580. https://doi.org/10.29106/fesa.1359358
15. Coşkun, E., & Sayılgan, G. (2008). Finansal sıkıntının dolaylı maliyetleri: İMKB’de işlem gören şirketlerde bir uygulama. Gazi Üniversitesi İktisadi ve İdari Bilimler Fakültesi Dergisi, 10(3), 45-66. https://dergipark.org.tr/tr/pub/gaziuiibfd/article/301019
16. Custódio, C., Ferreira, M. A., & Garcia-Appendini, E. (2023). Indirect costs of financial distress. Review of Finance, 27(6), 2233-2270. https://doi.org/10.1093/rof/rfad014
17. Deng, S., Huang, X., Zhu, Y., Su, Z., Fu, Z., & Shimada, T. (2023). Stock index direction forecasting using an explainable eXtreme Gradient Boosting and investor sentiments. The North American Journal of Economics and Finance, 64, 101848. https://doi.org/10.1016/j.najef.2022.101848
18. Deng, S., Luo, Q., Zhu, Y., Ning, H., & Shimada, T. (2024). Financial risk forewarning with an interpretable ensemble learning approach: An empirical analysis based on Chinese listed companies. Pacific-Basin Finance Journal, 85, 102393. https://doi.org/10.1016/j.pacfin.2024.102393
19. Donker, H., Ng, A., & Shao, P. (2020). Borrower distress and the efficiency of relationship banking. Journal of Banking & Finance, 112. https://doi.org/10.1016/j.jbankfin.2017.12.013
20. Fawcett, T. (2006). An introduction to ROC analysis. Pattern Recognition Letters, 27(8), 861-874. https://doi.org/10.1016/j.patrec.2005.10.010
21. Jones, S., Johnstone, D., & Wilson, R. (2017). Predicting corporate bankruptcy: An evaluation of alternative statistical frameworks. Journal of Business Finance & Accounting, 44(1-2), 3-34. https://doi.org/10.1111/jbfa.12218
22. Kaygın, C. Y., Tazegül, A., & Yazarkan, H. (2016). Estimation capability of financial failures and successes of enterprises using data mining and logistic regression analysis. Ege Academic Review, 16(1), 147-159. https://izlik.org/JA66BB98YH
23. Kulalı, İ. (2014). Muhasebe temelli tahmin modelleri ışığında, finansal sıkıntı ve iflasın karşılaştırılması. Sosyoekonomi, 22(22). https://doi.org/10.17233/se.81747
24. Lundberg, S. M., & Lee, S.-I. (2017). A unified approach to interpreting model predictions. Advances in Neural Information Processing Systems, 30, 4765–4774. https://proceedings.neurips.cc/paper/7062-a-unified-approach-to-interpreting-model-predictions
25. Ohlson, J. A. (1980). Financial ratios and the probabilistic prediction of bankruptcy. Journal of Accounting Research, 18(1), 109-131. https://doi.org/10.2307/2490395
26. Opler, T. C., & Titman, S. (1994). Financial distress and corporate performance. The Journal of Finance, 49(3), 1015-1040. https://doi.org/10.2307/2329214
27. Papana, A., & Spyridou, A. (2020). Bankruptcy prediction: The case of the Greek market. Forecasting, 2(4), 505-525. https://doi.org/10.3390/forecast2040027
28. Saito, T., & Rehmsmeier, M. (2015). The precision-recall plot is more informative than the ROC plot when evaluating binary classifiers on imbalanced datasets. PLOS ONE, 10(3), e0118432. https://doi.org/10.1371/journal.pone.0118432
29. Scavuzzo, C. M., Scavuzzo, J. M., Campero, M. N., Anegagrie, M., Aramendia, A. A., Benito, A., & Periago, V. (2022). Feature importance: Opening a soil-transmitted helminth machine learning model via SHAP. Infectious Disease Modelling, 7(1), 262-276. https://doi.org/10.1016/j.idm.2022.01.004
30. Shumway, T. (2001). Forecasting bankruptcy more accurately: A simple hazard model. The Journal of Business, 74(1), 101-124. https://doi.org/10.1086/209665
31. Tran, K. L., Le, H. A., Nguyen, T. H., & Nguyen, D. T. (2022). Explainable machine learning for financial distress prediction: Evidence from Vietnam. Data, 7(11), 160. https://doi.org/10.3390/data7110160
32. Tsai, C.-F., Hsu, Y.-F., & Yen, D. C. (2014). A comparative study of classifier ensembles for bankruptcy prediction. Applied Soft Computing, 24, 977-984. https://doi.org/10.1016/j.asoc.2014.08.047
33. Veganzones, D., & Séverin, E. (2018). An investigation of bankruptcy prediction in imbalanced datasets. Decision Support Systems, 112, 111-124. https://doi.org/10.1016/j.dss.2018.06.011
34. Zeytinoglu, E., & Akarım, Y. (2013). Financial failure prediction using financial ratios: An empirical application on Istanbul Stock Exchange. Journal of Applied Finance & Banking, 3, 107-115. https://ideas.repec.org/a/spt/apfiba/v3y2013i3f3_3_8.html
35. Zhang, Z., Wu, C., Qu, S., & Chen, X. (2022). An explainable artificial intelligence approach for financial distress prediction. Information Processing & Management, 59(4), 102988. https://doi.org/10.1016/j.ipm.2022.102988
36. Zmijewski, M. E. (1984). Methodological issues related to the estimation of financial distress prediction models. Journal of Accounting Research, 22, 59-82. https://doi.org/10.2307/2490859