Machine Learning-Driven Prediction of Offshore Vessel Detention: The Role of Neural Networks in Port State Control

This study investigates the application of different neural network (NN) models in assessing the risk of the detention of offshore vessels during port state control (PSC) inspections. The focus is on the use of different NN models (“nnet”, “mlp”, “neuralnet”, “rsnns”) to identify the main risk factors based on historical data on vessels and their inspections. The main objective of this research is to improve maritime safety and the efficiency of inspection procedures by applying techniques that can more accurately predict the probability of detention of the offshore vessels. These models make it possible to analyse complex patterns in the data, such as the relationships between the country of inspection, flag, memorandum, age, tonnage and previous deficiencies, and the risk of detention. Understanding these patterns is crucial for inspection teams’ proactive action as it helps direct resources to potentially high-risk vessels. Implementing these models into PSC processes helps to optimise resource allocation, reduce unnecessary costs, and increase the reliability of decision-making processes. NN models significantly help in recognising non-linear patterns and provide high accuracy in risk prediction. The study also includes a comparative analysis of the elements that determine the accuracy, sensitivity, and other performance aspects of the models to determine the most appropriate approach for practical implementation. The results emphasise the importance of applying artificial intelligence (AI) in various aspects of modern maritime safety management. This research opens up new opportunities for the development of intelligent support systems that not only increase safety but also improve the efficiency of inspection processes on a global scale.