Using Machine Learning Methods in Financial Distress Prediction: Sample of Small and Medium Sized Enterprises Operating in Turkey

Abstract

Financial distress has become one of the main topics on which lots of research has been done in the recent finance literature. This paper aims to predict the financial distress of Turkish small and medium firms using Logistic Regression, Decision Tree, Random Forest, Support Vector Machines, K-Nearest Neighbor and Naive Bayes model. Empirical results indicate that decision tree model is the best classifier with overall accuracy of %90 and %97 respectively for 1 and 2 years prior to financial distress. Three years prior to financial distress, Naive Bayes outperform other models with an overall accuracy of 92.86%. Furthermore, this study finds that distressed firms have more bank loans and lower equity. In the Turkish economy, where cyclical fluctuations are high in the last decade, distressed firms grew rapidly with high bank loans and gained higher operating profits than non-distressed firms. After a while, distressed firms that cannot manage their financial expenses get into financial trouble and go bankrupt. This article can be useful for managers, investors and creditors as well as its contribution to academic research.