Analysis of ship accidents in Turkish territories: A data mining application

Sea transport, where approximately 85% of the world trade volume is transported, is an event that is likely to be exposed to danger at any time due to the variability of its conditions. Although there are national/international studies, technological developments and innovations to keep these risks at a minimum, the situation remains uncertain. Sea traffic increased considerably when the speed of operation in loading and unloading ports is added to the increasing sea trade fleet. In an environment where 24 hours of activity is, never completed, marine accidents are inevitably formed due to the unknown variables. It has been seen in the past years that even the slightest negativity can have very big consequences. In the slightest sense, accidents that result in material damage, sometimes reach the big environmental disasters and even human's dead. For this reason, accurate assessment and evaluation of marine accident analyzes are of great importance in order to raise awareness by identifying the causes of accidents and making policies to develop preventive measures. It is seen that human error is the leading cause of accidents in almost all of the studies conducted in the world about marine accidents and statistical analyses. In the near future, even in a world where unmanned ships are in service, the human element will continue to exist. In today's information age where almost all kinds of data are recorded, the records kept in the maritime sector are also increasing day by day. The main subject of this study is to analyze the marine accidents by using the data mining method, which enables us to make meaningful and descriptive comments with computer programs by reaching meaningful information within large data heaps. The universe of the study is determined as sea accidents in Turkish territorial waters. In this context, the purpose of this study is to determine which variables work together in marine accidents by means of the association rule, which is one of the important analysis methods of data mining. Besides, the reasons of accident and accident type variables from data codes were analyzed by clustering and classification methods of data mining and it was tried to determine which variables were related to other variables. However, the data collected were interpreted and analyzed. Because of the analyses carried out, accidents have occurred usually in the ships, which has been loaded, ship at sailing, when there is no marine pilot in the ship, the vessels that do not have bow thruster, in the calm weather conditions. The ships are mostly foreign flags and they are dry/bulk cargo vessels. It is noteworthy that accidents occurred in Marmara sea and Istanbul strait region where regional and international ship traffic is more intense than other regions. In the outcomes of the accident association results; there have been no injuries, deaths, losses, marine pollution and vessel damage established, vessels did not request tug assistances.