Predicting the financial success and failure: Comparison of the logit regression and support vector machine

Financial distress is a phenomenon with socio-economic consequences that negatively affects firms, lenders and investors. The prediction of financial distress is important not only for the lenders but also for other participants of the society as well. Especially, the early warning models giving information about the possible financial failure cases are useful to take precautionary measures and to reduce the losses. This study compares the success of the Logistic Regression (LR) model and Support Vector Machine (SVM) model to predict the financial distress of firms. For this purpose, financial ratios of a total of 75 companies, 26 of which financially unsuccessful and 49 of which financially successful, that are traded in Borsa Istanbul (BIST) were calculated by using the financial statements between 2006-2017. In the light of these data, financial distress estimations were made separately from 1, 2 and 3 years before the financial failure date. According to the obtained results, the predictive success of the SVM model was higher than that of the traditional LR model which is valid for each of the 3 years. In addition, in this study, the ratios such as return on total assets, EBITDA profit margin, return on equity, cash ratio are found to be important in predicting the financial failure among 22 financial ratios used in the estimation of financial failure for these samples.