Link prediction of the global container shipping network using multi-index coupled learning based on random forest classifier
摘要
This study investigates the evolution of the global container shipping network through a multi-index link prediction framework. Simulations of adding high-probability predicted links reveal improvements in network connectivity, clustering and accessibility. Regional analysis reveals spatial heterogeneity in the distribution of potential links. These potential links are primarily concentrated within intraregional connections in Europe and East Asia, reflecting the demand for regional economic integration and the deepening of short-sea shipping networks. Interregional links are also prominent, especially between East Asia and Southeast Asia, Europe and West Asia, Europe and Africa, North America and Central America, East Asia and North America, and East Asia and Oceania. These patterns align with broader macro-trends, including global supply chain restructuring, nearshoring, and the ongoing importance of major trans-oceanic trade routes. The findings offer valuable insights into the evolution of global container shipping network and provide practical guidance for shipping companies and port authorities in route planning, service deployment, and long-term infrastructure strategy.