From Data to Security: Machine Learning Approaches for Maritime Piracy Risk Assessment

The main purpose of this research was to investigate the factors that most strongly affect the occurrence of maritime piracy incidents within the period of 2015–2024. In order to achieve this, statistical data were collected, aggregated, and analyzed with the aim of identifying variables that significantly contribute to the risk of pirate attacks at sea. These variables were then used as essential input for the development of a predictive model constructed with the use of artificial intelligence (AI) and machine learning (ML) techniques. The dataset covered several important dimensions, including the geographical regions where incidents occurred, the classification of different attack types, the nature and degree of violence used, and the tools and methods employed by the perpetrators.

To establish the relative importance of these features, advanced algorithmic approaches to feature selection were applied, with particular emphasis on classifiers based on the Random Forest method. This technique allowed the identification of variables that exert the greatest influence on the likelihood of future piracy incidents. The results demonstrated that three factors stand out as having the strongest impact: geographical location (with Southeast Asia and Africa highlighted as the most vulnerable regions), the type of attack (especially cases involving boarding of vessels), and direct violence against crews, including hostage-taking and kidnappings carried out for ransom.

The outcomes of this study provide a solid basis for the design and implementation of AI-powered early warning systems. Such tools can play an important role for a wide range of stakeholders, including shipping companies, governmental agencies, and international organizations concerned with maritime safety. By integrating predictive analytics into maritime security strategies, it becomes possible to monitor risks dynamically, recognize potential threats in real time, plan safer navigational routes, and allocate protective resources in a more efficient and cost-effective manner. Ultimately, this research highlights the potential of combining data-driven methods with AI to strengthen maritime security and reduce the human and economic costs associated with piracy.