By Using Logarithmic Regression and Artificial Neural Network to Improve Prediction Model of Dead Number Resulted from Road Traffic Accidents in Turkey

Ömer Faruk Cansız


Traffic accidents occurred on highway in Turkey cause materially and morally damage. To decrease the damage, prediction model developed. In this study, demographic and traffic data which from 1970 to 2007 are used. These data are consist of dependent and independent variables. Dependent variable is formed Number of Dead (ND). As for independent variables are comprised Population (P), Registered Number of Vehicle (VN), Vehicle-km (VK), Number of Drivers (DN). Models are developed using Artificial Neural Network (ANN) and Logarithmic Regression (LR) enhanced by Smeed. PVNVKDN model developed taking real values logarithm is the best performance of models in LR technique. VKDN created by using historical data sets is the best model in ANN technique. As for models created by randomly selected data, the best model is VKDN. When performances of best models are compared, VKDN is the best model because of lowest error rate.


Logarithmic Regression; Artificial Neural Network; Traffic Accident

Full Text:

PDF (Türkçe)


© 2011 Karaelmas Fen ve Mühendislik Dergisi