Comprehensive analysis of ship detention probabilities using binary logistic regression method with machine learning

Port State Control (PSC) is a regulatory process conducted by one or more qualified PSC officers (PSCOs) to ensure compliance with maritime safety and environmental standards. When significant deficiencies that pose substantial risks to maritime safety or the environment are identified during these inspections, the vessel may be detained. Nevertheless, the audit process is still frequently susceptible to human factors that have the potential to impact the outcome. This study aims to develop a predictive model to determine the probability of a ship being detained following an inspection. The dataset utilized in this study comprises 16.533 inspection reports from 2023, all within the Paris Memorandum of Understanding (Paris MoU) region. A binary logistic regression model was constructed using Python programming language to estimate the probability of detention. Before model construction, the Synthetic Minority Over-sampling Technique (SMOTE) was applied to address class imbalance during the data preparation phase. This model enables shipowners to predict the probability of their ships being detained in advance and make improvements pre-emptively. Consequently, time, cost, and reputational losses can be minimized. Furthermore, port states can enhance the efficiency of their processes by conducting more detailed or expedited inspections based on the model's results. This approach will reduce the influence of subjective human factors in the decision-making process, thereby facilitating the formation of more objective judgments.