Bankrupcy Risk Forecast Based on Company Balance Sheet Data Using Machine Learning Makine Ogrenmesi Kullanilarak Sirket Bilan^o Verilerine Dayali Iflas Riski Tahmini

Kesgin T., Shakeri S., Bulut N., Yüzük S., AKTAŞ M. S.

4th International Conference on Computer Science and Engineering, UBMK 2019, Samsun, Türkiye, 11 - 15 Eylül 2019, ss.195-200 identifier identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Doi Numarası: 10.1109/ubmk.2019.8907029
  • Basıldığı Şehir: Samsun
  • Basıldığı Ülke: Türkiye
  • Sayfa Sayıları: ss.195-200
  • Anahtar Kelimeler: Ohlson O-score, Altman Z-score, Bankrupcy Risk Assessment, Feature Extraction, Machine Learning
  • Yıldız Teknik Üniversitesi Adresli: Evet

Özet

Estimating the risk of bankruptcy of a company has significant economic impact for its owners, investors. In order to determine the risk of bankruptcy or, in other words, success or failure of the company in the future, different models are presented in the literature. The two prominent ones are the Ohlson O-score and the Altman Z-score models. Instead of evaluating all the balance sheets published in the past, these models estimate the bankruptcy risk, taking into account the latest published balance sheet, ie local data rather than historical data. The purpose of this study is to evaluate the risk of bankruptcy risk of the companies, the effect of the performance of the company in the past years, taking into account the effect of more successful estimates can be done. Therefore, the machine learning model has been formed and the predicted success of the model has been investigated on the balance sheet data, which is the label of the company on the risk of bankruptcy in previous periods. In the pre-processing stage of the model, the Information Gain and Principle Component Analysis approaches for the selection of attributes; Logistic Regression, Support Vector Machine and Random Forest are used as machine learning algorithm. The results revealed that a model learning from the previous balance sheet data estimates more successful bankruptcy risk than the financial models that decide on the basis of local data.