Revealing spatiotemporal connections in container hub ports under adverse events through link prediction

Abstract

Frequent adverse events have significantly impacted international trade. They disrupt the safety and stability of the global container shipping networks. To uncover potential connections among container hub ports, the K-shell and degree of node (KSDN) denoising algorithm denoises the liner hub-and-spoke shipping network. Based on both local and global information, the number of neighbors and the proportion of information transmitted and closeness (NNPITC) link prediction algorithm aims to achieve higher accuracy and faster computation speed. The NNPITC link prediction algorithm is compared with the other five directed weighted link prediction algorithms using Precision, Recall, F-measure, and Area Under the receiver-operating characteristic Curve (AUC) as evaluation metrics. The experimental results show that the NNPITC link prediction algorithm achieves the highest AUC value of 0.98624 among all the algorithms, demonstrating superior performance. The high-performance NNPITC link prediction algorithm is used to mine the potential connection relations among container hub ports from 2021 to 2023. Evolutionary trends in liner hub-and-spoke shipping network are explored. It provides valuable references for port shipping stakeholders to enhance the transshipment efficiency and risk resilience of liner hub-and-spoke shipping network.