科技创新与应用2025，Vol.15Issue(30)：50-54,5.DOI:10.19981/j.CN23-1581/G3.2025.30.013
一种时空融合自适应低通多图卷积神经网络
摘要
Abstract
This paper proposes a spatiotemporal fusion adaptive low-pass multi-graph convolutional network(STFALMGCN)for traffic flow prediction.First,we propose a causal gating linear unit module to extract the temporal characteristics of traffic flow,use causal convolution to greatly reduce the training parameters,and the gating linear unit speeds up the training speed.Secondly,a dynamically learned adaptive adjacency matrix is constructed using multiple information and based on the spatial information given by the graph structure to build a spatial graph,and global information and hidden information are fully utilized to find the optimal correlation graph.Finally,temporal and spatial features are integrated to further improve prediction accuracy.Experiments were conducted on two real-world datasets,and the experimental results showed that our model performed better prediction accuracy.关键词
交通流量预测/图卷积神经网络/深度学习/智慧城市/性能提升Key words
traffic flow prediction/graph convolutional neural network/deep learning/smart city/performance improvement分类
交通运输引用本文复制引用
茆浩,王超,杨青通,石剑锋,戚晨溪..一种时空融合自适应低通多图卷积神经网络[J].科技创新与应用,2025,15(30):50-54,5.基金项目
南通职业大学校级科研项目(23ZK14) （23ZK14）
