Abstract:

Traffic accidents occurred in Turkey caused their financial/moral losses to continue to maintain its status by virtue of being first during the agenda. Traffic accidents are the result of a combination of many factors. These can be listed as human, road, vehicle, climate and environmental conditions. As a result of traffic accidents, there may be accidents that can be recovered or accidents that cannot be compensated. In order to minimize the number and effects of traffic accidents, the factors causing the accident in general should be identified and eliminated. In order to identify the factors that cause traffic accidents accident history data are utilized. The important thing in accident analysis is the correct classification of the existing situation with the help of the model. As a result of methods with high classification success, the determination of the variables causing the accident can be determined more accurately. When literature studies for traffic accidents are examined, it is seen that discriminant analysis, logistic regression analysis and logarithmic linear models are generally used. In this study, 3181 traffic accidents, the result of which were fatal, injured in Antalya and its districts between 2012 and 2016, were discussed. In the study, the classification of traffic accidents has been made by using the machine learning methods, the usage of which has increased recently. The performances of the machine learning methods discussed were compared according to various criteria. Naive Bayes is the method that classifies accident data with the highest accuracy among machine learning methods.

Keywords: