Short-Term Traffic Flow Prediction Method for Urban Road Sections Based on Space–Time Analysis and GRU

Guowen Dai,Xuecai Xu,Changxi Ma

doi:10.1109/access.2019.2941280

Abstract

Accurate short-term traffic forecasts help people choose transportation and travel time. Through the query data, many models for traffic flow prediction have neglected the temporal and spatial correlation of traffic flow, so that the prediction accuracy is limited by the accuracy of traffic data. This paper proposed a short-term traffic flow prediction model that combined the spatio-temporal analysis with a Gated Recurrent Unit (GRU). In the proposed prediction model, firstly, time correlation analysis and spatial correlation analysis were performed on the collected traffic flow data, and then the spatiotemporal feature selection algorithm was employed to define the optimal input time interval and spatial data volume. At the same time, the selected traffic flow data were extracted from the actual traffic flow data and converted into a two-dimensional matrix with spatio-temporal traffic flow information. The GRU was used to process the spatio-temporal feature information of the internal traffic flow of the matrix to achieve the purpose of prediction. Finally, the prediction results obtained by the proposed model were compared with the actual traffic flow data to verify the effectiveness of the model. The model proposed in this paper was compared with the convolutional neural network (CNN) model and the GRU model, and the results show that the proposed method outperforms both in accuracy and stability.

Highlights

Accurate and real-time short-term traffic flow prediction can correctly and reasonably infer traffic conditions in the future according to the current traffic network change rules, provide convenient route planning for travelers, alleviate traffic congestion and reduce air pollution
In order to evaluate the effectiveness of the proposed method, two other models are used for comparative experiments [30]–[38], Convolutional Neural Network (CNN) and Gated Recurrent Unit (GRU) models
This study firstly analyzes the correlation of input traffic data, initializes the relevant time lag and the number of spatial segments used for prediction according to the selection rules, and constructs a neural network prediction model using the initialized network data

Summary

INTRODUCTION

Accurate and real-time short-term traffic flow prediction can correctly and reasonably infer traffic conditions in the future according to the current traffic network change rules, provide convenient route planning for travelers, alleviate traffic congestion and reduce air pollution. Cheng et al [9] quantitatively analyzed the temporal and spatial correlation of traffic flow, and established a hybrid process neural network model to achieve short-term traffic flow prediction. Dong and Shao [14] combined with the spatio-temporal characteristics of traffic flow, taking traffic flow under free-flow state as the research object, and establishing a prediction model based on state space to realize traffic flow prediction of multiple sections of road network. Tian and Pan [19] discussed the performance of the LSTM cyclic neural network for predicting short-term traffic flow and compared it with several other commonly used models. Based on the summary of previous studies, this study integrates spacetime analysis with gated loop unit neural network algorithm, applies in the field of short-term traffic flow prediction, and conducts in-depth research and discussion.

SPATIAL CORRELATION ANALYSIS OF TRAFFIC FLOW DATA

EXPERIMENTAL STEPS AND RESULTS

EXPERIMENTAL STEPS

CONCLUSION AND PROSPECT