Predicting Density in Merge and Diverge Areas Based on their Geometry and Traffic Characteristics using Metaheuristic Procedure

In this paper, it was attempted to predict the density (which is a common criterion which represents the quality of flow in every freeway segment) in the merge and diverge areas by simulating 2880 different merge and diverge areas with different geometry and different traffic characteristics. Density was obtained for each merge and diverge areas after analyzing trajectory data. A database containing density as a function and geometric and traffic characteristics as its variables was generated after determination of densities in all instances. By using this database, two models were developed by Artificial Neural Network (ANN) and Particle Swarm Optimization (PSO) algorithm to predict density in these areas. The models were tested, validated and their errors were checked. The results indicated a good accuracy of similarity between the results of models in predicting density and that of simulations. Case studies were surveyed to verify the accuracy and validity of models. Statistical analysis showed that there were no significant differences between means of densities predicted by models and surveyed densities from case studies.