WAFQ - Maritime and Port Analytics

Edge-cloud networks are essential for enabling intelligent applications in maritime transport, ports, and global supply chains, where energy efficiency and rapid task execution are critical requirements. However, existing approaches often rely on static resource information, limiting their adaptability to the dynamic and online nature of these deployments. This raises significant challenges in optimizing resource utilization while ensuring high performance. To address these issues, we introduce GreenMaritimeEdge (GME), a novel framework designed to tackle the "green workload coordination problem" in maritime systems. GME optimizes the coordination of model partitioning and function assignment using Reinforcement Learning with Human Feedback (RLHF), leveraging a human-in-the-loop approach to enhance decision-making in complex environments. Extensive evaluations in real-world maritime applications, including large international ports and intricate supply chain scenarios, demonstrate that GME achieves superior energy efficiency and task execution speed compared to traditional methods. By dynamically adapting to changing conditions, GME provides a robust solution for managing intelligent maritime applications, addressing both environmental and operational challenges. This study highlights the transformative potential of GME to improve sustainability and efficiency in maritime transport and global supply chain networks, paving the way for future advancements in intelligent edge-cloud solutions.

This paper aims to identify the key characteristics influencing whether container ports in Northern Europe and the Mediterranean function as gateways, pure transshipment hubs, or mixed ports combining both roles. While previous research has explored the conditions favoring container hub activities, a quantitative assessment of how specific variables contribute to a port's transshipment profile remains absent. In this study, these variables are grouped into three factors: Port Operations and Infrastructure, Hinterland and Logistics, and Attractiveness to Shipping Lines. A Multinomial Logistic Regression (MRL) model was employed, revealing that Port Operations and Infrastructure is the only significant factor differentiating gateway ports from mixed ports. In contrast, the distinction between gateway ports and pure transshipment hubs is driven by Hinterland and Logistics, as well as Attractiveness to Shipping Lines, while Port Operations and Infrastructure proves statistically insignificant. Additionally, the model correctly classified approximately 85% of the 65 ports in the sample, with greater accuracy in the gateway and pure transshipment categories. Understanding the critical factors driving transshipment port success is crucial for governments and terminal operators to avoid common pitfalls, particularly the so-called "hubsession"-an overemphasis on becoming a carrier hub without adequately addressing fundamental determinants.

This study aims to analyze the global shipping patterns of vehicle carriers on a port-specific basis by utilizing the AIS vessel movement database. The significance and connectivity of ports in maritime transportation networks are evaluated using network centrality metrics, including Closeness Centrality, Betweenness Centrality, and Eigenvector Centrality. Gephi visualization software is employed for network analysis, allowing for a detailed comparison of centrality results across the three main companies to identify and compare their key ports. This research provides a comprehensive understanding of the global vehicle carrier shipping network and highlights the critical hubs driving its efficiency and connectivity.

Freight rate fluctuations drastically influence the profitability of shipowners, with cascading impacts across the maritime value chain. While considerable research has explored machine learning (ML) applications for freight rate forecasting, there is no comprehensive resource that consolidates both proven ML methodologies and the influential factors driving these models. Addressing this gap, this study systematically reviews 28 papers published between 2012 and 2024, compiling 78 independent variables – such as crude oil prices, fleet size, and newbuilding prices – commonly used as inputs for ML models. Neural Networks dominate the landscape of existing research, although hybrid and specialized models often outperform standalone approaches. The review further highlights a strong focus on the dry bulk market (61% of studies), while the prevalent data periodicity is weekly (42.9%) or daily (39.3%). Accuracy metrics such as RMSE, MAPE, MAE, and MSE are most frequently applied. By consolidating these methodologies and input factors into a single source, this paper serves as a valuable reference for advancing predictive accuracy and supporting strategic decision-making in maritime logistics.