Modeling of Driver Lane Choice Behavior with Artificial Neural Networks (ANN) and Linear Regression (LR) Analysis on Deformed Roads

Abstract

In parallel to the economic developments, the importance of road transportation was significantly increased in Turkey. As a result of this, long-distance freight transportation gains more importance and hence numbers of the heavy vehicles were significantly increased. Consequently, road surface deformations are observed on the roads as the increasing freight transportation and climatic conditions influence the road surface. Therefore, loss of functionality of the road surface is observed and drivers are much prone to accident due to their driving characteristics as they can have more tendencies to change their lanes not to pass through the deformation area. In this study, the lane changing behaviors of the drivers were investigated and both Artificial Neural Network (ANN) and Linear Regression (LR) models were proposed to simulate the driver behavior of lane changing who approach to a specific road deformation area. The potential of ANN model for simulating the driver behavior was evaluated with successive comparison of the model performances with LR model. While there was a slight performance increase for the ANN model with respect to LR model, it is quite evident that, ANN models can play an important role for predicting the driver behavior approaching a road surface deformation. It can be said that, approaching speed plays an important factor on the lane changing behavior of a driver. This can be criticized by the fact that, drivers with high approaching speeds more likely pass through the deformation to avoid the accidents while changing their lanes with a high speed.