Spatial-Temporal Aggregation Graph Convolution Network for Efficient Mobile Cellular Traffic Prediction

Accurate cellular traffic prediction is challenging due to the complex spatial topology of cellular network and the dynamic temporal feature of mobile traffic. To overcome these problems, this letter proposes a spatial-temporal aggregation graph convolution network (STAGCN), in which the daily histo...

Full description

Saved in:
Bibliographic Details
Published in:IEEE communications letters 2022-03, Vol.26 (3), p.587-591
Main Authors: Zhao, Nan, Wu, Aonan, Pei, Yiyang, Liang, Ying-Chang, Niyato, Dusit
Format: Article
Language:English
Subjects:
Agglomeration
Cellular communication
Convolution
Correlation
deep learning
Feature extraction
graph convolution network
Mobile traffic prediction
Predictive models
Research and development
spatial-temporal correlation
Time series analysis
Topology
Traffic models
Urban areas
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Accurate cellular traffic prediction is challenging due to the complex spatial topology of cellular network and the dynamic temporal feature of mobile traffic. To overcome these problems, this letter proposes a spatial-temporal aggregation graph convolution network (STAGCN), in which the daily historical pattern and the hourly current-day pattern of mobile traffic are modeled. Moreover, the complex spatial-temporal correlation is captured by an aggregation graph convolution network for all nodes across different timestamps. The external factors' impact on mobile traffic is fed into a regression module at the last step to obtain the predicted traffic. Experimental results show that the proposed model can achieve better prediction performance than conventional methods with superior training efficiency.
ISSN:1089-7798
1558-2558
DOI:10.1109/LCOMM.2021.3138075