Empirical Formulation of Highway Traffic Flow Prediction Objective Function Based on Network Topology

Abstract

Accurate Highway road predictions are necessary for timely decision making by the transport authorities. In this paper, we propose a traffic flow objective function for a highway road prediction model. The bi-directional flow function of individual roads is reported considering the net inflows and outflows by a topological breakdown of the highway network. Further, we optimise and compare the proposed objective function for constraints involved using stacked long short-term memory (LSTM) based recurrent neural network machine learning model considering different loss functions and training optimisation strategies. Finally, we report the best fitting machine learning model parameters for the proposed flow objective function for better prediction accuracy.