A Multiscale-Grid-Based Stacked Bidirectional GRU Neural Network Model for Predicting Traffic Speeds of Urban Expressways

Abstract

In recent decades, studies on short-term traffic speed forecasting of the large-scale road are a new challenge for researchers and engineers. Especially based on deep learning neural networks, studies on short-term traffic forecasting have achieved mush-room growth. This study proposes a stacked Bidirectional Gated Recurrent Unit neural network model to predict the traffic speed of the expressway over different estimation time intervals in an effective manner. By building a multiscale-grid model, it can take less time to derive a set of key traffic parameters of different scales to predict traffic speed of the various-scale road. The speed prediction of small-scale sections can cover more detailed road spatial features preparing for Vehicle Navigation System, and the speed prediction of large-scale sections can establish the real-time traffic control strategies. In order to validate the effectiveness of the proposed model, we use the floating car data, with an updating frequency of 1 minute from the urban freeway of Beijing, for model training and testing. The experimental results show that the stacked BiGRU network with the multiscale-grid model enables to capture the spatial-temporal characteristics of traffic speed efficiently. Furthermore, the BiGRU with two layers (BiGRU-2L) outperforms benchmark models in the prediction of the traffic speed, which presents a significant advantage in reducing the overfitting problem, decreasing the excessive time-consuming and improving the effective use of limited computation resources.