FCEP - Maritime and Port Analytics

Predicting Vessel Traffic Flow (VTF) is a critical component of intelligent maritime transportation management, using Automatic Identification System (AIS) data analysis. Traditional deep learning methods often fail to fully capture the complex and dynamic nature of VTF data. To overcome these limitations, this study introduces a novel approach: the Multi-view Periodic-Temporal Network with Semantic Representation (MPTNSR). This model integrates three perspectives-periodic, temporal, and semantic-to provide a comprehensive solution for VTF prediction. The periodic and temporal features of VTF, often hidden in its evolution, are effectively extracted using a hybrid Convolutional Neural Network and Bidirectional Long Short-Term Memory framework. Meanwhile, in real-world applications where multiple target regions are involved, the semantic view leverages a Graph Convolutional Network (GCN) to capture correlations in VTF fluctuations across regions based on geographical and statistical similarities. By fusing insights from all three views, the model achieves accurate prediction. Additionally, an optimised loss function incorporating both local and global metrics ensures robust performance. Extensive comparative experiment results demonstrate that the MPTNSR model significantly outperforms fourteen methods in VTF prediction tasks. Its ability to deliver precise prediction makes it a valuable tool for port planning, enhancing safety and sustainability in maritime transportation.

Accurate forecasting of ship order book is of vital importance in the maritime industry, as it enables companies to optimize operations, allocate resources efficiently, and support strategic decision-making. However, the inherent volatility and dynamic nature of historical ship order book data present significant challenges in achieving accurate, intelligent, and robust predictions. This study introduces a novel ensemble deep random vector functional link (edRVFL) algorithm designed to predict future trends of ship order books. The proposed edRVFL model combines deep feature extraction with ensemble learning techniques to improve forecasting accuracy. Additionally, we implement a discontinuous and parsimonious embedding strategy, which diverges from the traditional continuous time step approach employed in standard edRVFL models. This parsimonious embedding method not only reduces model complexity but also enhances generalization performance. Extensive experiments on ship order book datasets demonstrate the effectiveness of the proposed approach, with comparative analyses confirming its superiority over existing forecasting methods. The developed edRVFL model offers a reliable and efficient solution for ship order book prediction, contributing significantly to the maritime industry's operational planning and decision-making processes.

Container availability, be it empty or laden, affects cargo stuffing, container loading, vessel space utilisation, and order fulfilment. To fully cater for the rapid changes in slot demand, ship liners do not only require a simultaneous optimisation of equipment supply and laden container allocation. They also need to diversify their equipment supplies to enable more flexible equipment planning and therefore more efficient slot allocation on vessels. Based on a previously developed three-echelon collaborative slot allocation planning model which investigated the dynamics among local, regional, and global scales, with containers loading and discharge at multiple vessels and service loops, and another empty container repositioning yield-based optimisation model with uncertain demand, this paper proceeds to investigate scenarios of shipping equipment demand and supply mismatch cases, which have already been optimised in their slot allocation and empty container repositioning. For this sake, a novel integrated mixed-integer programming model for slot allocation with comprehensive consideration of the container supply planning under container/slot demand uncertainty is developed. An Long-short term memory (LSTM) neural network prediction mechanism is developed and integrated with the optimisation process. Extra benefits of this enhanced methodology is reviewed.

The shipping market plays a crucial role in the global economy, while the ship building and recycling processes can lead to significant environmental impacts. To mitigate these issues and promote sustainable practices within the maritime industry, it is important to align the construction of new ships and their subsequent recycling with market demands. This study aims to conduct a macro-level analysis to examine and forecast the volumes of ships newbuilding and recycling, where automatic ARIMA models are adopted for time series prediction. The modeling results indicate that the Root Mean Squared Error (RMSE) ranges from 0.08 to 1.19, and the Median Absolute Percentage Error (MDAPE) varies from 0.31% to 4.94%, which are considered as satisfactory. We predict the gross tonnage of ship newbuilding and recycling until the year 2033. The results demonstrate the shifting dynamics in the global shipbuilding and recycling industries, with policy implications for strategic adjustments to maintain competitiveness and achieve sustainable maritime practices. In future work, we will further conduct geo-spatial analysis and explore more advanced or comparative models to better predict the shipbuilding and recycling volumes.