Abstract

The Korean shipping industry's vulnerability to economic crises, such as the 2016 bankruptcy of Hanjin Shipping, highlights the need for robust bankruptcy prediction methods, particularly for small and medium-sized enterprises (SMEs). This research aims to identify key risk factors for predicting bankruptcy in shipping companies from Korean industry by leveraging financial, non-financial, and economic data through advanced machine learning models, including Extreme Gradient Boosting (XGBoost) and Long Short-Term Memory (LSTM) networks. A comprehensive literature review and interviews with industry practitioners were conducted to refine the variables used in the models. These models predict bankruptcy across 1, 3, and 5-year horizons, with Explainable Artificial Intelligence (XAI) techniques employed to interpret the impact of each variable. The findings reveal that non-financial and macroeconomic variables, such as LIBOR interest rates and trade volume growth rates, are significant predictors across all periods. Additionally, the importance of financial ratios, especially those related to profitability, increases with the length of the forecasting period. The research also highlights distinct risk factors between large shipping companies and SMEs, underscoring the need for tailored risk management strategies. This study contributes valuable insights for stakeholders in the shipping market, including policymakers and financial institutions. The identified risk factors enable shipping companies to improve strategic planning and anticipate market cycles more effectively. Policymakers can use these insights to develop regulations that address the unique needs of shipping SMEs. Overall, this research provides a comprehensive understanding of the market dynamics, offering practical implications for managing bankruptcy risk in the shipping industry. Further approach can be held with wider geographical areas to reflect specific regional aspects with much advanced prediction models.

Keywords

shipping, Bankruptcy prediction, Machine learning

Document Type

Thesis

Publication Date

2024

Embargo Period

2024-07-31

Download

Share

COinS