Abstract
Effective maritime accident prediction will benefit both maritime safety management and the insurance industry. Due to the complex non-linearity and non-stationarity nature of maritime accident data, its prediction is still a challenge in the research field. An autoregressive integrated moving average with explanatory variables (ARIMAX) model was proposed to predict maritime accidents accurately, and a multi-factor accident prediction framework was developed. Additionally, the impacts of eight influencing factors on the number of maritime accidents were also investigated, and the predictions from the ARIMAX model were contrasted with those from earlier maritime accident prediction models, as well as autoregressive integrated moving average (ARIMA), back-propagation neural network (BPNN), and support vector regression (SVR). The findings imply that an increase in any one of the eight factors may increase the number of maritime accidents worldwide. The ARIMAX model, which incorporates accident factors, is accurate enough to estimate the number of global maritime accidents and outperforms the ARIMA, BPNN, and SVR models in terms of prediction precision and robustness. The ARIMAX model outperforms earlier marine accident prediction models and has good applicability.
Acknowledgment
This work was financially supported by the National Natural Science Foundation of China (grant number50909058), and the National Key R&D Program of China (grant number 2021YFC2801005).
Disclosure statement
No potential conflict of interest was reported by the author(s).
Additional information
Funding
Notes on contributors
Jinhui Wang
Dr. Jinhui Wang, an associate professor at Shanghai Maritime University, is engaged in research in accident prevention and analysis, statistics of shipping accident, risk assessment.
Shaogang Zhang
Dr. Shaogang Zhang, a lecturer at Shanghai Maritime University,is engaged in shipping and marine engineering safety.