MFDGCN: Multi-Stage Spatio-Temporal Fusion Diffusion Graph Convolutional Network for Traffic Prediction

Abstract

Traffic prediction is a popular research topic in the field of Intelligent Transportation System (ITS), as it can allocate resources more reasonably, relieve traffic congestion, and improve road traffic efficiency. Graph neural networks are widely used in traffic prediction because they are good at dealing with complex nonlinear structures. Existing traffic prediction studies use distance-based graphs to represent spatial relationships, which ignores the deep connections between non-adjacent spatio-temporal information. The use of a simple approach to fuse spatio-temporal information is not conducive to obtaining long-term deep spatio-temporal dependencies. Therefore, we propose a new deep learning model Multi-Stage Spatio-Temporal Fusion Diffusion Graph Convolutional Network (MFDGCN). It generates multiple static and dynamic spatio-temporal association graphs to enhance features and adopts the multi-stage hybrid spatio-temporal fusion method. This promotes the effective fusion of a spatio-temporal multimodal and uses the diffuse convolution method to model the graph structure and time series in traffic prediction, respectively. The model can better predict both long and short-term traffic simultaneously. We evaluated MFDGCN using real road network traffic data and it shows good performance.

ADSTGCN: A Dynamic Adaptive Deeper Spatio-Temporal Graph Convolutional Network for Multi-Step Traffic Forecasting

“…Adjacency matrix : The connectivity among all sensors in the traffic network can be depicted by matrix , commonly referred to as the adjacency matrix, . In our work, the connectivity of edges in the graph is represented using the distance and similarity between nodes [ 37 ].…”

“…Using distances between sensors to create graph adjacencies tends to ignore richer spatial relationships. This paper uses the multi-association graph method in [ 37 ] to create graph networks that extract rich spatial dependencies. Spatial static graph represents the neighborhood spatial structure of the traffic network, which is generated based on the distance between road sensors.…”

A Hybrid Model for Spatiotemporal Air Quality Prediction Based on Interpretable Neural Networks and a Graph Neural Network

“…The input for self-attention is the feature matrix output obtained from splicing several city characteristics that were retrieved using multiple INNs. Natural language processing frequently uses encode-decode to manage mapping relationships and to map one sequence onto another more effectively [59,60]. The GNN module receives its input from the self-attention output.…”

Spatio-Temporal Traffic Flow Prediction Based on Coordinated Attention

“…At this stage, much literature, including [3][4][5][6][7], explore the complex nonlinear relationships of traffic flow from both time and space dimensions. A convolutional neural network is the main tool for traffic flow spatial information exploration.…”

“…A traffic prediction model with traditional convolutional network as the main body can only extract local spatial information based on the limitation of grid data input, which cannot fully reflect the traffic variation of a complete road network, so graph convolutional networks gradually appear in the study of traffic problems, which we will introduce in detail in the following sections. With the continuous development of deep learning techniques, attention mechanisms are used to dynamically capture the long-term dependence of spatio-temporal information of traffic flow, based on the variability of traffic information in different spaces over time, for example, the traffic distribution in commercial and residential areas is not the same in the morning, midday and evening, and the potential correlation between such spatio-temporal information of traffic flow is often ignored, such as ASTGCN [4], IGAGCN [5], and MFDGCN [6] and other models capture the spatial dependence and temporal dependence of traffic information through two independent components, ignoring the dependence under the potential association of spatio-temporal information, and not specifying the degree of influence of multiple traffic flow characteristics in different spatio-temporal dimensions.…”