|
Spatiotemporal Traffic Prediction Using Graph Neural Networks, Recurrent Units, and the Spatial Structure of the Network
|
Fateme Noori Alamouti *   |
| K.N Toosi university of technology |
|
|
Abstract: (25 Views) |
Optimal urban traffic management requires accurate and timely prediction of roadway conditions at spatiotemporal scales, as congestion patterns in road networks are influenced by complex spatial interactions among streets and dynamic variations over short- and long-term time horizons. Therefore, developing approaches capable of simultaneously extracting multidimensional dependencies across space and time plays a key role in improving traffic management efficiency and urban decision-making. In this study, using spatial data from the road networks of four districts in Tehran and traffic classes extracted from Google images over a three-month period, a hybrid framework based on Graph Neural Networks (GNNs) and Recurrent Units is proposed for traffic level prediction. The road network is modeled as a directed graph and used as input to the graph neural network. To enhance the spatial representation of the model, first- and second-order adjacency matrices, as well as matrices of centrality measures—including betweenness, closeness, and weighted degree—are incorporated as inputs to the GNN.The GNN extracts spatial relationships and interactions among road segments, having the results of GNN along with temporal sequences and traffic classes as inputs to the GRU model, traffic predictions are generated. This model captures the spatial and temporal dependencies of traffic classes within the road network by leveraging enriched spatial data and historical traffic information. The results demonstrate that the proposed framework can more effectively learn complex spatiotemporal traffic patterns and achieve higher prediction accuracy compared to baseline models. Complementary analyses indicate that this improvement is primarily due to the inclusion of topological and structural indices, highlighting the significant role of the spatial structure of the network in traffic analysis and prediction.
|
|
| Keywords: Traffic prediction, spatiotemporal models, network spatial structure, graph neural networks, recurrent units |
|
|
|
Type of Study: Research |
Subject:
GIS
Received: 2026/01/11 | Accepted: 2026/06/16 | ePublished ahead of print: 2026/06/17
|
|
|
|
|
|
|
| Send email to the article author |
|
|
