Analysis and estimation of traffic accident results in Kahramanmaraş with machine learning methods

Traffic accidents are one of the biggest causes of human injuries and deaths. Every year, more than 1 million deaths and 50 million injuries occur in traffic accidents around the world. While there is a traffic accident every 2 seconds in these accidents, it results in a fatal accident every 50 seconds. In this study, 7929 traffic accidents, consisting of 28 features, that took place in Kahramanmaraş city center and its districts between 2017 and 2020 were analyzed by machine learning techniques in the WEKA tool. Classification, clustering, association rule and visual comparison methods are used in the analysis of traffic accidents. Naive Bayes, Support Vector Machine, K Nearest Neighbor, PART, AdaboostM1, J48 and Random Forest algorithms were used in the classification method. SimpleKMeans and EM clustering algorithms were used in the clustering method. Apriori algorithms were used in the association method. As a result, when the performance criteria of the classification algorithms in the prediction of traffic accidents are examined, the J48 algorithm from the decision trees obtained the most successful results, while the success level of the NB algorithm was lower than the other algorithms, although it was not bad. The common result of all methods is that potential accidents with death and injury that may occur in the future are mostly between 04:00 and 08:00 in the morning and between 16:00 and 20:00 in the evening, in the settlement area of Kahramanmaraş city center and Onikisubat district, on slopes outside the settlement in Göksun district. It has been determined that the risk of living is high on winding and winding roads and on divided roads outside the settlement in Pazarcık district. In addition, it was determined that the drivers did not comply with the traffic and speed rules in the accidents that occurred, and the road, environment and safety equipment such as traffic lights, traffic signs, pedestrian ways, safety lanes and banquettes were insufficient.