Bayes' Theorem based maritime safety information classifier

• Published in: Chinese Control and Decision Conference pp. 2725 - 2729
• Main Authors: Liu, Hongze, Liu, Zhengjiang, Wang, Xin, Cai, Yao
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.06.2018
• Subjects: Classifier; e-Navigation; Lenses; Marine vehicles; Maritime Safety Information; Mathematical model; Naive Bayes; Navigation; Navigational Warning; Probability; Safety; Training
• ISSN: 1948-9447
• DOI: 10.1109/CCDC.2018.8407588

In order to ensure the safe navigation of ships, the Global Maritime Distress and Safety System was incorpo­rated into the Safety of Life at Sea regulations, and the amendment has already came into force. Became mandatory for compulsorily fitted craft, the system includes an important Maritime Safety Information Broadcasting System. Nowa­days, due to the lack of relevant norms, the seafarers in many sea areas still need to read and identify the theme of the messages artificially, with low efficiency and accuracy. To solve this problem, several machine learning based solutions have been presented and analyzed. However, a mathematically solidly founded naive Bayes classifier did now show its best during the previous research. Thus, the Bayes' Theorem based classifier is explored in detail. Navigational Talex dataset, with thousands of messages collected in navigational area VI from 2011 to 2016 are used, and naive Bayes classifiers with different event models, tokenization algorithms, and word bag sizes are compared in terms of accuracy, precision and recall rate, and F-Measure respectively. The chosen classifier performs well on the newest Navigational Talex messages.