Maritime safety through multi-source data fusion: an AdaBoost-based approach for predictive ship detention by port state control

Abstract

In global maritime safety, the efficiency of Port State Control (PSC) is paramount in ensuring the safety of sea. Facing the challenge of effectively identifying high-risk vessels, this study innovatively enhances PSC inspection efficiency. This study aims to reduce maritime accidents by significantly improving the efficiency and accuracy of detained vessels prediction through multi-source data fusion technology and the enhanced AdaBoost algorithm. AdaBoost, improves model accuracy by combining multiple weak classifiers. By comprehensively analyzing ship inspection records from 2015 to 2022 across various Chinese ports, combined with additional vessel information, this research constructs a developed predictive model to forecast the likelihood of ships being detained by PSC in Chinese ports. The proposed model successfully identified numerous non-detained and detained ships and achieved highly satisfactory predictive results on the training dataset. Through in-depth analysis of crucial evaluation metrics such as precision, recall, F1 score, and ROC, this study provides strong technical support for accurately identifying high-risk vessels that play a vital role in enhancing maritime safety. Moreover, our findings offer valuable insights for port managers to optimize ship selection processes and for shipping companies to improve operational efficiency, having a profound impact on the safety and development of the maritime transportation industry.