PREDICTING FINANCIAL FAILURE: EMPIRICAL EVIDENCE FROM PUBLICLY-QUOTED FIRMS IN DEVELOPED AND DEVELOPING COUNTRIES

This paper analyzes the data of 570 firms from developed and developing countries between 2010 and 2019 in an attempt to create high-accuracy financial failure prediction models. In this sense, we utilize three different methods, namely logistic regression (LR), artificial neural networks (ANN), and decision trees (DT), Aand compare the classification accuracy performances of these techniques. Using 16 financial ratios as independent variables, ANN is able to generate the most accurate prediction and outperforms the other methods in predicting failure. Otherwise said, ANN yields a correct classification accuracy of 98.1% one year prior to failure while LR and DT achieve accuracy rates of 94.7% and 96.1%, respectively. Furthermore, the empirical results demonstrate that the classification accuracy rate reaches 92.5% by ANN, 91.1% by DT, and 84.4% by logistic regression two years in advance. The findings of current research provide valuable insights into financial failure prediction and may entice practical implications for stakeholders, especially investors and regulatory bodies, by indicating that the use of the ANN approach mayAbe more effective.