Machine learning applications for risk assessment in maritime transport: Current status and future directions

As the maritime transportation system is inherently complex and vulnerable to potential hazards, it is critical to conduct risk assessments via applicable methodologies. Recently, machine learning (ML) algorithms have attracted tremendous attention due to their ability to analyze risks effectively. Nevertheless, there is a lack of a systematic summarization of ML applications in maritime transport risk assessment (MTRA). Hence, this review aims to encapsulate the current status, issues, considerations, and future directions of ML applications using the systematic reviews and meta-analyses method. In particular, the status is summarized from the following three dimensions: advantages, disadvantages, and corresponding applications. Moreover, several issues are recognized, including dataset processing and methods utilization, and considerations from the perspective of sensitivity analysis and evaluation methods. Regarding future directions, promising opportunities in terms of data and method improvements are identified. Overall, this review contributes to MTRA by presenting the existing research status with a framework and providing suggestions on model selection and method improvement for future research.