Short-Term Prediction of Traffic Speed using Recurrent Neural Networks (RNN)

Traffic forecasting, as an important part of intelligent transportation systems, plays an important role in monitoring traffic conditions. Given that many studies have done traffic prediction work with deep learning models, no research has been done on traffic speed prediction in different chapters. Considering the important effects of spatio-temporal factors and the excellent performance of recursive neural networks in the field of time series analysis, in this paper, one of the deep learning neural networks that combine the characteristics of Giti recursive neural networks in consecutive time data is proposed. The data used in this study are obtained from the active visualization and evaluation network in Seattle, USA. This research has compared three seasons of spring, summer and autumn and four deep learning models including LSTM, GRU, ConvLSTM and BiLSTM and a shallow SVM model in three time steps of 5 minutes, 10 minutes and 15 minutes. The results of this research show that the results of the proposed model There was no significant difference in different seasons, and also the four deep learning models and the SVM model did not have significant differences in different seasons and these differences can be ignored. The other results show that as the interval of time steps increases, the errors increase and the accuracy of the model decreases. According to the obtained results, the accuracy of the FI-GRU model is 0.52% higher than the lowest accuracy of the deep learning model (BiLSTM), and the accuracy of the proposed model(SVM) is 1.24% higher than the shallow learning model, as well as the proposed model for predicting traffic speed. It has performed 44.1% better in the time step of 5 minutes.