머신러닝을 활용한 해운기업 파산 예측에 관한 연구

Abstract

Purpose: This study aims to enhance bankruptcy prediction accuracy in the shipping industry by using advanced machine learning models, specifically XGBoost and LSTM, comparing the predictive factors between large enterprises and SMEs. Research design, data, and methodology: Utilizing financial data from 2001 to 2023 for the Korean shipping industry, the study analyzes key financial ratios and macroeconomic indicators. XGBoost and LSTM models are employed to develop customized bankruptcy prediction models for large enterprises and SMEs. Results: The XGBoost model outperforms others, effectively handling complex financial and macroeconomic variables. Large enterprises are more influenced by internal financial factors, while SMEs are sensitive to external economic conditions like a shipping index. Conclusions: The study highlights the need for tailored prediction models to improve bankruptcy risk management in the shipping industry, especially focusing on the size of the company. Future research should explore model applicability across different regions, and should consider integrating real-time data for enhanced accuracy.