Analysis of injury traffic accidents with machine learning methods: Adana case


Ozden C., ACI Ç.

PAMUKKALE UNIVERSITY JOURNAL OF ENGINEERING SCIENCES-PAMUKKALE UNIVERSITESI MUHENDISLIK BILIMLERI DERGISI, cilt.24, sa.2, ss.266-275, 2018 (ESCI) identifier identifier

Özet

In this study, a dataset is created using numeric data of injury traffic accidents in monthly base between 2005 and 2014 years in Adana province and meteorological data of the same years in order to develop prediction models which estimate number of traffic accidents involving injury and number of injured people. Feedforward Multilayer Artificial Neural Network, Function Fitting Artificial Neural Network, Generalized Regression Artificial Neural Network, Regression Tree, Support Vector Machine and Multiple Linear Regression Analysis methods were used in the prediction models. As a result of the study, SVM gives the most successful results for both prediction scenarios. Prediction of the number of traffic accidents involving injury is more successful than prediction of number of injured people except Regression Tree method. In addition, it has concluded that it is possible to take precautions using road and weather data of the accidents which occurred in previous years.