A deep spatiotemporal approach in maritime accident prediction: A case study of the territorial sea of South Korea

Abstract

Predicting the risk of maritime accidents is crucial for improving traffic surveillance and marine safety. The availability of data sources and development of machine learning and deep learning methodologies can improve operational risk prediction. Similar to larger vessels, numerous ocean accidents are caused by small- and medium-sized vessels owing to poor conditions and defective equipment. This study aims at investigating the application of deep learning in both short- and long-term predictions of different types of accident risks associated with small vessels by considering multiple influencing factors. Herein, several big data sources that contain data collected from the territorial sea of South Korea, including ocean accidents, ocean depth, weather data, and small vessel trajectories, are utilized. Four machine learning and six deep learning algorithms are implemented and compared in nine scenarios with three different grid sizes using daily, weekly, and monthly models. The results reveal that although the performance of the proposed deep spatiotemporal ocean accident prediction (DSTOAP) model varies according to grid sizes and time intervals, its accuracy (more than 78%) makes it reliable for predicting accidents. Furthermore, although all types of accidents are captured with high accuracy, more than 84% of collision accidents can be predicted accurately. For practical applications, the results of this study can guide ocean accident management and safety planners in choosing appropriate methods for different time schedules and grid sizes, according to the range of coverage of the patrol ship.