Öznitelik Seçim Yöntemleri ve Makine Öğrenmesi Kullanarak Şirket Bilanço Verilerine Dayalı İflas Riski Tahmini

Öz

Bir şirketin başarısı hem firmanın iç muhatapları hem de yatırımcılar ve üçüncü kişilerce büyük önem taşımaktadır. Finansal olarak başarısızlık kimi zaman iflaslar ile sonuçlanabilmekte ve firmanın muhatapları üzerinde yıkıcı etkiler yaratabilmektedir. Yatırımcılar, finansörler, yöneticiler bazen de politika yapıcıları için firmaların iflas risklerini tahmin etmek oldukça önemlidir. Literatürde iflas riskinin tahmini için birçok yöntem geliştirilse de Ohlson O-skoru ve Altman Z-skoru iflas riskini tahmin için oldukça sık kullanılan iki yöntemdir. Bu iki modelin hem lineer model olmaları hem de firmaların yalnızca son bilançolarıyla ilgilenmeleri bazen hatalı tahminlere yol açabilmektedir. İflas olgusunun bir süreç olduğu düşünüldüğünde şirketin sadece son finansal raporlarının incelenmesi bir takım sakıncalar barındırır. Bu sebeple iflas risklerini doğru tahmin etmek için şirketlerin geçmiş finansal raporlarının da incelenmesi gerekmektedir. Literatürdeki bu iki iflas riski tahmin yöntemi şirketlerin sadece son finansal raporlarıyla ilgilenmektedir. Ayrıca bu iki modelde şirketin başarısına dair karar verilemeyen gri alanlar bulunmaktadır. Bu çalışmada literatürdeki klasik lineer modeller yerine, lineer olmayan makine öğrenmesi algoritmaları kullanılarak şirketlerin iflas riskleri tahmin edilmeye çalışılmıştır. Bu amaç doğrultusunda öznitelik seçim metodu olarak Bilgi Kazanımı ve Temel Bileşenler Analizi, Lineer Diskriminant Analizi ile birleştirilerek ve makine öğrenmesi metodu olarak Lojistik Regresyon, Karar Destek Vektörleri ve Rassal Orman algoritması kullanılmıştır. Bu bağlamda şirketlerin iflas riskini makine öğrenmesi algoritmalarıyla tahmin etmenin, lineer klasik modellerden başarılı olduğu sonucuna ulaşılmıştır.

Anahtar Kelimeler

• Ohlson O-skoru,Altman Z-Skoru,Şirket İflas Riski Tahmini,Öznitelik Çıkarımı,Makine Öğrenmesi

Bankruptcy Risk Forecasting Based on Company Balance Sheet Data Using Feature Selection Methods And Machine Learning

Öz

The success of a company has a significant issue for both the interlocutors of companies and other related persons. Financial failure sometimes end up bankrupt and can have a critical effect on the company’s interlocutors. Prediction of bankruptcy is significant for investors, backers, directors and sometimes policymakers. Although there are a lot of models to predict bankruptcy in the financial literature, Ohlson O-score and Altman Z-score are models that are used quite often. The fact that these two models are both linear models and companies are only interested in their latest balance sheets can sometimes lead to incorrect predictions. Considering bankruptcy as a process, to interest in only the latest financial reports of the companies has some drawbacks. For this reason, in addition to the current, previously financial reports of companies should be interested to predict bankruptcy risk of the company correctly. In the literature, these two classical models interest in only the current financial reports of companies. Additionally, there are grey areas that are not decided about the bankruptcy of companies in these two classical models.   In this study, it is tried to predict the bankruptcy risk of companies by using non-linear machine learning algorithms rather than classical linear models in the financial literature. In line with this main purpose, as feature selection methods Information Gain, Principle Component Analysis algorithms by combining Linear Discrimination Analysis algorithm and as machine learning methods Logistic Regression, Support Vector Machine, and Random Forest algorithms are used. It has been found that predicting the bankruptcy risk of companies by using non-linear machine learning algorithms is more successful than linear classical models.

Anahtar Kelimeler

• Ohlson O-score,Altman Z-score,Bankruptcy Risk Prediction,Feature Extraction,Machine Learning

Kaynakça

1. [1] Ohlson, James A. "Financial ratios and the probabilistic prediction of bankruptcy." Journal of accounting research ,1980, pp. 109-131
2. [2] Altman, Edward I. "Financial ratios, discriminant analysis and the prediction of corporate bankruptcy." The journal of finance 23.4 , 1968, pp. 589-609
3. [3] Taffler, Richard J. "Empirical models for the monitoring of UK corporations." Journal of Banking & Finance 8.2, 1984, pp. 199-227
4. [4] Imelda, Elsa, and Ignacia Alodia. "The Analysis of Altman Model and Ohlson Model in Predicting Financial Distress of Manufacturing Companies in the Indonesia Stock Exchange." Indian-Pacific Journal of Accounting and Finance 1.1, 2017, pp. 51-63
5. [5] Wang, Nanxi. "Bankruptcy prediction using machine learning." Journal of Mathematical Finance 7.04, 2017, pp. 908
6. [6] Altaş, Dilek, and Selay Giray. "Mali Başarısızlığın Çok Değişkenli İstatistiksel Yöntemlerle Belirlenmesi: Tekstil Sektörü Örneği.", 2005
7. [7] Muzir, Erol and Caglar, Nazan (2009), “The Accuracy of Financial Distress Prediction Models in Turkey: A Comparative Investigation with Simple Model Proposals”, Anadolu University Journal of Social Sciences, 9(2): 15-48
8. [8] Raut, Sneha, Milind Tiwari, and Kuldeep Kumar. "Financial Distress Prediction Using Cutting-Edge Statistical Techniques: A Study of Australian Real Estate Sector." Shodh-Amrit: JKLU Journal of Engineering & Management 1.2, 2018, pp. 2-32

9. [9] Viviani, Jean‐Laurent, and Hong LE Hanh. "Why Do Banks Fail?-The Explanation from Text Analytics Technique.", SSRN Electronic Journal, 2018
10. [10] Lin, Wei‐Chao, Yu‐Hsin Lu, and Chih‐Fong Tsai. "Feature selection in single and ensemble learning‐based bankruptcy prediction models." Expert Systems 36.1, 2019, e12335
11. [11] Vu, Loan Thi, et al. "Feature selection methods and sampling techniques to financial distress prediction for Vietnamese listed companies." Investment Management & Financial Innovations16.1, 2019, pp. 276
12. [12] AKPINAR, Onur, and Gökçe AKPINAR. "Finansal Başarısızlık Riskinin Belirleyicileri: Borsa İstanbul’da Bir Uygulama.", Journal of Business Research – Turk, Volume 9, Issue 4 2017
13. [13] Zeytinoglu, Emin, and Yasemin Deniz Akarim. "Financial failure prediction using financial ratios: An empirical application on Istanbul Stock Exchange." Journal of Applied Finance and Banking 3.3, 2013, pp. 107
14. [14] Bellovary, Jodi L., Don E. Giacomino, and Michael D. Akers. "A review of bankruptcy prediction studies: 1930 to present." Journal of Financial education, 2007, pp. 1-42
15. [15] Acosta-González, Eduardo, Fernando Fernández-Rodríguez, and Hicham Ganga. "Predicting corporate financial failure using macroeconomic variables and accounting data." Computational Economics 53.1, 2019, pp. 227-257
16. [16] Chen, James Ming. "Models for Predicting Business Bankruptcies and Their Application to Banking and to Financial Regulation." Available at SSRN 3329147, 2019
17. [17] Shakeri, S., & Ashouraei, M. (2016). NeurAda: Combining artificial neural network and Adaboost for accurate object detection. International Journal of Next-Generation Computing, 7(2), pp. 155-163
18. [18] Christopoulos, Apostolos G., et al. "An implementation of Soft Set Theory in the Variables Selection Process for Corporate Failure Prediction Models. Evidence from NASDAQ Listed Firms." Bulletin of Applied Economics 6.1, 2019, pp. 1-20
19. [19] Fernández, Manuel Ángel, et al. "Focused vs unfocused models for bankruptcy prediction: Empirical evidence for Spain." Contaduría y Administración 64.2, 2019, e96
20. [20] Varatharajan, R., Manogaran, G., & Priyan, M. K. (2018). A big data classification approach using LDA with an enhanced SVM method for ECG signals in cloud computing. Multimedia Tools and Applications, 77(8), 10195-10215
21. [21] Zengin T., Shakeri S., Bulut N., Yüzük1 S. and Aktaş M.S. (2019, September). Bankruptcy Risk Forecast Based on Company Balance Sheet Data Using Machine Learning. In 2019 4rd International Conference on Computer Science and Engineering (UBMK). IEEE. 2019