Özet:

In our daily life, transportation has an important place. Many people use land transport and go from one point to another by their own vehicle. Some of the dangerous movements that people have done during the day (i.e. acceleration, sudden slowdown, sudden reverse to the right, sudden reverse to the left, sudden shift of the strip, etc.)  As a result, accidents occur. These dangerous movements conducted by drivers can be modeled and driver profiles can be created. According to the driver’s created profiles, the drivers may be warned or the necessary sanctions may be imposed so that they do not do such behaviors that may cause an accident during driving. In this study, the driver’s driving data was taken during the use of the vehicle and modeled the driver’s behavior with various machine learning algorithms (i.e. acceleration, sudden reverse to right, sudden reverse to left). As a result of this modeling, the success rates of machine learning algorithms were 65.50 for the random forest, 47.97 for the Bayes networks, 59.55 for the decision tables, 55.84% for the artificial nerve networks and 53.82% for the support vector machines. The windowing method was used to increase success in the classification of driver behaviors and achieved 89.61% success rates in the random forest, 90.90% in the Bayes networks, 92.20% in the decision tables, 84.41% in the artificial nerve networks, and 90.90% in the support vector machines. In the modeling of driver behavior, the windowing method creates a positive impact while the highest success rate of machine learning algorithm is defined as Decision tables.

Anahtar Kelimeler: