Predicting driver's decision- making behaviour in Amber phase using ML techniques

Generally drivers face a dilemma as they approach the intersection during the amber phase. Due to the existence of this Dilemma zone, safety and efficiency of the intersection affect. Whereas, decision-making behaviour depends upon different parameters such as approaching speed, vehicular volume per cycle, type of vehicle, distance from stop line, number of lanes at the intersection, yellow phase and driver's attributes such as age and gender. The two main contributions offered by this paper are first, developing the prediction and classification models of driver's decision using Artificial Neural Network (ANN) and Support Vector Machine (SVM). Second, defining the importance of parameters using Random Forest which influences the driver's decision-making behaviour. For this study, 328 driver's decision or responses were collected through video graphic survey conducted at three different locations of Hyderabad, India. The research concludes that SVM with the sigmoidal kernel is showing more classification accuracy when compared with other kernels. Whereas; when SVM (71.95%) and ANN (76.82%) models are compared than ANN was found to be having more accuracy. It was found that distance from stop-line, approaching speedand driver's age is found to the most affecting parameters among all considered parameters.