Seaport Resilience Analysis and Throughput Forecast Using a Deep Learning Approach: A Case Study of Busan Port

Abstract

The global nature of seaport operations makes shipping companies susceptible to potential impacts. Sustainability requires seaport authorities to understand the underlying mechanisms of resilience in a dynamic world, to ensure high performance under disruptions. This paper deals with data analytics for analysing port resilience and a new paradigm for productivity forecasting that utilize a hybrid deep learning method. Nonlinear analytical methods include Lyapunov exponent, entropy analysis, Hurst exponent, and historical event analysis, with statistical significance tests. These approaches have been utilised to show that throughput demand at Busan port (South Korea) exhibits complex behaviour due to business volatility. A new forecasting method based on long short-term memory (LSTM) and random forest (RF) has been applied to explore port throughput in realizing recovery policy. The LSTM networks have shown high effectiveness in time-series forecasting tasks; RF is proposed as a complementary method to mitigate residual errors from the LSTM scheme. Statistical significance tests have been conducted to comprehensively evaluate the introduced forecasting models. The results show that the hybrid method outperformed three benchmarked models in both the short- and long-term forecasting at a 95% confidence level, guaranteeing accuracy and robustness as well as suitability. As a seeking strategy for seaport competitiveness, novel resilience planning incorporates sustainability to prepare for disruptions such as a global pandemic.