University of Hertfordshire

From the same journal

By the same authors

Documents

View graph of relations
Original languageEnglish
JournalIEEE Access
Early online date21 Aug 2020
DOIs
Publication statusE-pub ahead of print - 21 Aug 2020

Abstract

Accurate and reliable traffic flow prediction is critical to the safe and stable deployment ofintelligent transportation systems. However, it is very challenging since the complex spatial and temporaldependence of traffic flows. Most existing works require the information of the traffic network structure andhuman intervention to model the spatial-temporal association of traffic data, resulting in low generality of themodel and unsatisfactory prediction performance. In this paper, we propose a general spatial-temporal graphattention based dynamic graph convolutional network (GAGCN) model to predict traffic flow. GAGCN usesthe graph attention networks to extract the spatial associations among nodes hidden in the traffic featuredata automatically which can be dynamically adjusted over time. And then the graph convolution networkis adjusted based on the spatial associations to extract the spatial features of the road network. Notably, theinformation of rode network structure and human intervention are not required in GAGCN. The forecastingaccuracy and the generality are evaluated with two real-world traffic datasets. Results indicate that ourGAGCN surpasses the state-of-the-art baselines

ID: 22503827