Study on the prediction model of accidents and incidents of cruise ship operation based on machine learning. (15th September 2022)

• Record Type: Journal Article
• Authors: Su, Zhaoqian; Wu, Cuilin; Xiao, Yingjie; He, Hongdi
• Is Part Of: Ocean engineering. Volume 260(2022)
• Publication Date: 2022-09-15
• Subjects: Cruise ship operation -- Accidents and incidents -- Machine learning -- Ensemble learning; Ocean engineering -- Periodicals; Ocean engineering; Periodicals; 620.4162
• DOI: 10.1016/j.oceaneng.2022.111954
• Languages: English
• ISSNs: 0029-8018

Abstract

Abstract: To improve the safety of cruise ship operations in China, machine learning was used to build an ensemble learning model to predict accidents and incidents during cruise ship operations. First, the characteristics of accidents and incidents in cruise ship operations were analysed. The types of accidents and incidents of cruise ship operations include not only loss of life and property but also early or delayed parking and the failure of passengers' recreational wishes caused by shaking due to waves and swells. The number of accidents and incidents of cruise ship operations is highly related to the number of passengers, luggage, gross tonnage, and voyage. Therefore, 10 machine learning algorithms were selected to train the data of accidents and incidents of cruise ship operations at the Shanghai Wusongkou International Cruise Port (SWICP). The predictive model of accidents and incidents of cruise ship operations was quantitatively evaluated using two performance indices: determination coefficient (r 2 ) and root mean square error (RMSE). Finally, an improved ensemble learning model (KNN + LR + ExtraTree) was proposed. The proposed improved model showed the best predictive performance compared with the other models in this study. Highlights: The types of accidents and incidents of cruise ship operations include not only loss of life and property but also early or delayed parking and the failure of passengers' recreational wishes caused by shaking due to waves and swells. The number of accidents and incidents are highly related to the number of passengers, luggage, gross tonnage and voyage. Therefore, 10 machine learning algorithms were selected to train the data of accidents and incidents of cruise ship operations at the Shanghai Wusongkou International Cruise Port (SWICP). The predictive model was quantitatively evaluated using two performance indices: determination coefficient (r 2 ) and root mean square error (RMSE). Finally, an improved ensemble learning model (KNN + LR + ExtraTree) was proposed.