Research On Traffic Information Analysis Technology Based On Graph Neural Networ

The significant role of traffic prediction in intelligent transportation systems(Intelligent Transportation Systems,ITS)has received increasing attention from researchers.Researchers have proposed many methods to improve traffic prediction performance.Due to complex temporal,spatial,and inherent long-term predictions,accurate traffic forecasting remains challenging,especially for long-term forecasting tasks.This paper studies traffic prediction algorithms based on graph neural networks.In this paper,three effective algorithms are proposed to solve the problem of traffic forecasting.The main research results of this article include the following aspects.(1)This paper designs a new deep learning method for traffic prediction tasks for long-term traffic flow prediction tasks,which is based on the Spatial-Temporal ChannelAttention based Graph Convolutional Network(STCAGCN).We design an attention mechanism to learn complex temporal and spatial correlations.Then we develop the stacked spatial-temporal convolution layer to model complex temporal and spatial correlations.Each spatial-temporal convolution layer is composed of a gated time convolution network and a graph convolution network.We develop a gated time convolution network to model non-linear temporal correlations,which process long sequences through stacked dilated convolution.Moreover,the graph convolution network exploits the hidden spatial correlations via learning the self-adaptive adjacency matrix.Experiment results on the real-world datasets demonstrate that the proposed STCAGCN model obtains improvements over the state-of-the-art.(2)This paper proposes a new deep learning framework called the Spatial-Temporal Multi-Feature Fusion Network(STMFFN)for short and short-term traffic flow prediction tasks.Specifically,a multi-scale attention module with temporal convolution is designed to capture the temporal dependencies from different scales.Then,a gated graph convolution module is proposed,which constructs adaptive adjacency matrices,and integrates graph convolution and graph aggregation modules to capture spatial dependencies from different ranges.Moreover,a multi-feature fusion layer is presented to fuse the extracted spatial and temporal dependencies by obtaining the attention vectors of temporal and spatial features.The experimental results of the real-world datasets show a continuous improvement of 6%-9% compared to the state-of-the-art baseline.(3)This paper proposes a novel Multi-scale temporal Attention based spatialtemporal Conditionally Parameterized Graph Convolution Network(MA-CPGCN)for temporal dependencies and large spatial dependencies tasks with different step sizes.Specifically,MA-CPGCN captures spatial,short-term,and long-term dependencies from different scales.In particular,multi-scale temporal attention is developed to jointly leverage self-attention and deep convolution to model short and long-range temporal features across multiple time steps.Furthermore,a conditional parameterized graph convolution network is proposed to dynamically model different spatial features.Experimental results on real-world datasets show that the proposed MA-CPGCN model achieves improvements over the most advanced models,especially in long-term traffic prediction.