Machine Learning Applications for Berth Allocation Optimization in Major Red Sea Ports: A Multi-Port Analysis

Mohammed H. Alshareef, Department of Supply Chain Management and Maritime Business, Faculty of Maritime Studies, King Abdulaziz University, Jeddah, Saudi ArabiaFollow

Abstract

Efficient berth allocation remains a critical challenge in Red Sea port operations, particularly given the diverse operational capabilities across the region's ports. The research utilized historical operational data from 147,892 berth allocation instances across these ports spanning 2019--2023. A deep learning model was developed using four dense layers with ReLU activation functions, incorporating vessel characteristics, weather patterns, and operational constraints. The model's performance was validated through comparison with traditional allocation methods using real-time port operation data. Port selection criteria included geographical distribution along the Red Sea, varying levels of technological adoption, and willingness to share operational data for research purposes. The machine learning model achieved a 33.2% reduction in vessel waiting time and a 27.7% improvement in berth utilization compared to conventional methods. Environmental consideration integration resulted in a 21.4% decrease in fuel consumption during berthing operations. This study develops and validates a machine learning-based approach to optimize berth allocation decisions while considering multiple operational constraints and environmental factors specific to the Red Sea region. The research focused on five major ports: Jeddah Islamic Port (Saudi Arabia), Port Sudan (Sudan), Ain Sokhna Port (Egypt), Aqaba Port (Jordan), and Port of Djibouti (Djibouti), selected based on their strategic location, cargo volume, and varying levels of operational efficiency. Furthermore, the model demonstrated 91.3% accuracy in predicting optimal berth assignments under varying operational conditions, with performance ranging from 93.5% in normal operations to 84.8% in combined stress scenarios. Performance improvements showed strong correlation with technological readiness (r = 0.94, p < 0.001), with more technologically advanced ports showing better optimization potential. The developed machine learning approach provides a robust framework for optimizing berth allocation in Red Sea ports, effectively balancing operational efficiency with environmental considerations.

First Page

Last Page

Recommended Citation

Alshareef, Mohammed H. (2025) "Machine Learning Applications for Berth Allocation Optimization in Major Red Sea Ports: A Multi-Port Analysis," Journal of King Abdulaziz University: Marine Science: Vol. 35: No. 1, Article 9.