Combining Two-Stage Cluster Analysis and Decision Tree Algorithm to Marine Accidents Identification Model

• Main Authors: Chien, Wei-Jen, 錢威任
• Other Authors: Huang, Tsan-Huang
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/273428

碩士 === 國立臺灣海洋大學 === 運輸科學系 === 105 === Port state control (PSC) has been performed for many years. Its main purpose is to enhance the safety of ship navigation and reduce the occurrence of marine accidents. However, the benefits of PSC remain unclear. The study first focused on analyzing PSC data and numbers of marine accidents to help decision makers prevent such accidents. A total of 6,942 datasets were obtained from the Transportation Safety Board (TSB) of Canada for the period from 2004 to 2015. These datasets were combined with Paris MoU annual report data. The evaluation variables were the accident ratio, deficiency ratio, and detention ratio of deficiencies in the basing of the flags of ships. Two-stage cluster analysis was conducted on the basing of 32 ship flags, and a decision tree algorithm was used to form the if-then semantic rules. The output may help to achieve higher security benefits for navigation. The complete database can be used for the future establishment of a database of flag states. The marine accidents identification model is a scalable design architecture that facilitates the operation of ship safety inspections for full navigation.