Machine Learning-Based Gap Acceptance Model for Uncontrolled Intersections Under Mixed Traffic Conditions

Abstract

Uncontrolled intersections are the most common type of intersections in a transportation network. The study modeled minor road driver’s decision of accepting or rejecting a gap at four-legged uncontrolled intersections having similar geometric characteristics using Artificial Neural Network (ANN) model, Logistic Regression (LR) model, and Support Vector Machine (SVM) model. The results reveal that the performance of LR and SVM models are somewhat similar, while the performance of ANN model exceeds the performance of both LR and SVM models with a correct prediction of about 96.2%. Also, the higher values of the goodness of fit measures like F1 score and R2 value together with a lower value of MSE show that ANN model is better in distinguishing between the classes. The variable gap duration has a major influence on model prediction comparing to other variables. The effect of the critical gap, occupancy time, conflicting volume, and vehicle type are also found remarkable. © 2023, Transportation Research Group of India.