.png){kind=icon}
.png){kind=icon}
Return
Adaptive Spatio–Temporal Feature Graph Convolutional Network Prediction Model With Edge Computing Integration
.png){kind=icon}
PRE
AI
DOI：10.1109/JIOT.2026.3661976.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Traffic flow prediction constitutes a critical component of intelligent transportation systems (ITSs), yet remains highly challenging due to complex spatio–temporal dynamics and substantial data processing requirements. To address these challenges, this article proposes an adaptive spatio–temporal feature graph convolutional network (ASTFGCN), an efficient traffic flow prediction model based on edge computing architecture that integrates the distributed processing capabilities of edge computing, the lightweight modeling advantages of an adaptive feature enhancement (AFE) mechanism, and the spatial feature extraction capabilities of graph convolutional networks (GCNs), thereby achieving efficient and accurate traffic flow prediction. First, ASTFGCN leverages edge computing architecture to distribute traffic data preprocessing and modeling tasks to edge nodes, effectively reducing computational burden on central servers, minimizing data transmission latency, and enhancing real-time processing capabilities. Concurrently, model training is performed on cloud servers while edge nodes execute predictions using pretrained parameters, achieving optimal resource allocation and task decomposition for improved prediction efficiency. Second, ASTFGCN incorporates an AFE mechanism that utilizes spatial and temporal embeddings to generate attention weights for fusing and weighting raw input features. This approach maintains model compactness while enhancing the expressiveness of salient features. In addition, a dynamic adjacency matrix generation method is designed to automatically adjust the graph structure based on current features, improving the model’s adaptability to dynamic variations in traffic networks. Furthermore, by integrating GCNs, ASTFGCN efficiently captures spatial dependencies within traffic networks and complex interaction patterns between nodes, thereby improving prediction accuracy. Experimental results on multiple real-world traffic datasets demonstrate that ASTFGCN achieves superior prediction accuracy and computational efficiency, exhibiting significant advantages particularly in resource-constrained edge computing scenarios. This research provides a novel and practical solution for efficient traffic flow prediction in ITSs.
Keywords:
Adaptive feature enhancement (AFE)
edge computing
graph convolutional networks (GCNs)
traffic volume prediction
Journal
IF:
8.9
Papers:
1.4W
Citations:
7.8W