Bankruptcy Prediction in Lithuania’s Food Service Sector: Application of Traditional, Machine Learning, and Hybrid Models and the Impact of Macroeconomic Factors

Abstract

This study examines bankruptcy prediction models specifically for Lithuania's food service sector, which is known for its high economic sensitivity and significant bankruptcy risk. The primary issue addressed is the limited accuracy of traditional bankruptcy prediction models when applied to this industry—an important concern for business management and investors. The aim of the study is to evaluate the effectiveness of different types of bankruptcy prediction models for companies in Lithuania’s food service sector and to develop a hybrid model based on advanced technologies, tailored to the specific characteristics of the sector. The models investigated include the Altman Z-score, Ohlson O-score, Support Vector Machine (SVM), and Gradient Boosting Machine (GBM), with macroeconomic indicators incorporated into the advanced models. The empirical analysis was conducted using a dataset comprising 96 Lithuanian food service companies. Model performance was assessed through metrics such as AUC (ROC), F1 score, Brier score, and other key indicators. The findings reveal that the developed hybrid model with integrated macroeconomic factors achieved the highest prediction accuracy at 93.06%, along with the best overall balance of sensitivity, precision, and calibration, highlighting its potential as an effective tool for practical bankruptcy risk assessment in this sector.