Estimation of a Financial Distress Prediction Model Based on the Integration of the Support Vector Machine Algorithm and the Least Squares Model

Abstract

Objective: The objective of this study is to propose a hybrid model based on Partial Least Squares (PLS) and Support Vector Machine (SVM) to predict corporate financial distress and enhance the accuracy and stability of the prediction process. Methodology: This study utilized a dataset of 120 companies, consisting of 56 bankrupt and 64 non-bankrupt firms, over a two-year period. Initially, financial data were analyzed, and key features were extracted using the Partial Least Squares (PLS) method. The Support Vector Machine (SVM) algorithm was then employed, utilizing a grid search technique with 5-fold cross-validation to optimize model parameters. The performance of the proposed model was compared with traditional methods such as logistic regression and artificial neural networks. Findings: Empirical results indicated that the hybrid PLS-SVM model achieved an accuracy rate of 87% on the test set, outperforming traditional models and other machine learning techniques. Additionally, the model successfully identified the most relevant financial indicators for predicting financial distress and determined the role of each variable in the prediction process. Conclusion: Due to its high accuracy, interpretability, and significant stability, the proposed model can serve as an effective tool for financial institutions in risk management, credit approval, and financial planning processes. This study demonstrates that combining machine learning methods can improve financial prediction capabilities.