ABSTRACT

When performing microscopic traffic simulations, a premise is to have appropriately calibrated parameter values. This is often computationally expensive as it requires repeatedly running simulations. In this paper, we propose a machine learning (ML) + particle swarm optimization (PSO)-based methodology for calibrating microscopic traffic simulator parameters to improve computational efficiency. We first develop ML models that input the parameters to predict simulation outputs. Four machine learning models: decision tree, support vector machine, Gaussian process regression, and artificial neural networks are considered. The best-performing model is then embedded in PSO to seek the set of parameters that minimizes the difference between the predicted simulation outputs and the field observations. The ML+PSO methodology is applied to TransModeler using field data in Shanghai, China. We find that artificial neural networks yield the best prediction accuracy. Furthermore, PSO with embedded artificial neural networks shows superior computational efficiency and effectiveness..