Financial Distress Prediction Using Artificial Neural Network, Partial Least Squares Regression, Support Vector Machine Hybrid Model, and Logit Model

Financial distress refers to the situation where a firm’s cash flows are insufficient to meet contractually required payments. This has caused concern among capital owners and compelled financial analysts to employ a variety of methods to assess companies’ equity and analyze the firm’s financial status. Assessing and predicting financial distress in a timely and accurate manner can aid decision-makers in finding the optimal solution and preventing it. Numerous models have been developed thus far to predict and evaluate financial distress. The prediction accuracy has been improved through the use of various innovative methods. Using financial ratios and market data as independent variables and obtaining patterns for the financial forecast is one of the most important methods for evaluating the financial stability of businesses. Therefore, the primary objective of this study is to evaluate the performance of five models in this field, compare their accuracy of prediction, and ultimately select the best model to predict financial distress for a specified period in Iran. Specifically, the logit model, artificial neural network (ANN), support vector machine (SVM), partial least squares regression (PLS), and a hybrid model of SVM and PLS were chosen, analyzed, and compared. The results of the average accuracy of prediction indicate that the SVM has the highest accuracy one year before the onset of financial distress. In addition, findings from the two years preceding the failure indicate that the SVM-PLS model provides the most accurate classification of financially distressed and non-distressed firms.

کلیدواژه ها:

Artificial Neural Networks ، Financial distress ، Hybrid Model ، Logit Model ، Support vector machine