Port cluster analysis and feature importance in the Guangdong-Hong Kong-Macao Greater Bay Area: an integrated machine learning approach

This study employs an integrated machine learning framework to analyze the cluster structure and feature importance of major ports in the Guangdong-Hong Kong-Macao Greater Bay Area (GBA). Key methodologies include hierarchical clustering for uncovering port relationships, random forests and Lasso regression for feature selection, and PCA and t-SNE for dimensionality reduction and visualization. The results highlight container throughput, cargo throughput, and GDP as critical factors influencing port competitiveness, alongside the emerging importance of innovation and foreign trade. These findings provide actionable insights for enhancing regional supply chain integration and optimizing resource allocation. The study offers practical recommendations for port managers and policymakers while paving the way for future research on advanced machine learning applications in maritime logistics.