مرکز منطقه ای اطلاع رساني علوم و فناوري - THE ANALYSIS OF ACCIDENTS IN THE CITY OF ANKARA: AN ARTIFICIAL NEURAL NETWORK MODEL

Abstract :

This study aims to develop an artificial neural network model to address one of the negative aspects of urbanization, namely, the ever-increasing traffic accidents. For that purpose a feed forward model of artificial neural network has been created for the specific case of Ankara. In the construction of this model, the reports of accidents that occurred within the provincial limits of Ankara during the 2008-2010 period have been taken as the basis for our project, a risk analysis has been conducted and the techniques of artificial intelligence have been used to determine the risky zones. These reports have been obtained from the General Directorate of Security. The model takes the weather, road and vehicle conditions into consideration and provides the driver with a list of risky points on the route to the point of destination; he/she is also allowed to choose an alternative route. In creating the model, R2013a freeware version of MATLAB software and especially the Levenberg-Marquardt algorithm has been used. Since it has been proven that this algorithm has a very high percentage of success, this type of algorithm has been chosen to train and test the data. Our results have shown 95% confidence level. As part of our research, various functions of Google Maps technology such as mapping, placing marker and route-calculating have also been used as a visual guidance interface in order to figure out the routes of minimum risk for the drivers.