Multi-Branch Traffic Flow Prediction Based on Temporal Speed

Abstract

AbstractEfficient traffic management is a major issue for developing countries. Traffic flow prediction is an important problem in intelligent transportation system (ITS). Various studies have been reported in the literature for traffic flow prediction in which combined models have been proposed only using traffic flow data. It is evident from the traffic flow theory that speed and flow are inter-related. Therefore, considering speed to predict flow in a model can help in improving the performance of a prediction method. Keeping this in mind, we propose a traffic flow prediction method consisting of two branches. First branch predicts traffic flow using past flow data through long short-term memory (LSTM) neural network. Second branch predicts volume using Gaussian process regression (GPR) based on temporal speed. Finally, prediction from both the branches was combined through weighted average. The mean squared error (MSE), root mean squared error (RMSE), and Pearson’s correlation coefficient (r) were used to evaluate the effectiveness of the proposed model. Based on these measures, it was found that results of our proposed model are promising.KeywordsMulti-branchGPRTraffic flowLSTM