Abstract:While considering the spatial and temporal features of traffic, capturing the impacts of various external factors on travel is an essential step towards achieving accurate traffic forecasting. However, existing studies seldom consider external factors or neglect the effect of the complex correlations among external factors on traffic. Intuitively, knowledge graphs can naturally describe these correlations. Since knowledge graphs and traffic networks are essentially heterogeneous networks, it is challenging to integrate the information in both networks. On this background, this study presents a knowledge representation-driven traffic forecasting method based on spatial-temporal graph convolutional networks. We first construct a knowledge graph for traffic forecasting and derive knowledge representations by a knowledge representation learning method named KR-EAR. Then, we propose the Knowledge Fusion Cell (KF-Cell) to combine the knowledge and traffic features as the input of a spatial-temporal graph convolutional backbone network. Experimental results on the real-world dataset show that our strategy enhances the forecasting performances of backbones at various prediction horizons. The ablation and perturbation analysis further verify the effectiveness and robustness of the proposed method. To the best of our knowledge, this is the first study that constructs and utilizes a knowledge graph to facilitate traffic forecasting; it also offers a promising direction to integrate external information and spatial-temporal information for traffic forecasting. The source code is available at <a class="link-external link-https" href="https://github.com/lehaifeng/T-GCN/tree/master/KST-GCN" rel="external noopener nofollow">this https URL</a>.

What problem does this paper attempt to address?

This paper attempts to solve several key problems in traffic prediction: 1. **Capture of spatio - temporal features**: Traffic flow has significant spatial and temporal correlations, that is, the traffic state of the current road section is not only affected by the historical traffic conditions of the upstream and downstream road sections, but also related to the historical traffic conditions of adjacent roads. Existing research has modeled these spatio - temporal features through methods such as Graph Convolutional Network (GCN) and Recurrent Neural Network (RNN). 2. **Influence of external factors**: In addition to spatio - temporal features, traffic flow will also be affected by a variety of external factors, such as weather conditions, the existence of traffic hubs, emergencies, holidays, and the distribution of surrounding Points of Interest (POI). However, most of the existing studies either ignore these external factors or do not fully consider the impact of their complex associations on traffic. 3. **Introduction of knowledge graph**: In order to better capture the semantic relationships between external factors and their impact on traffic, the author proposes to construct a Knowledge Graph and integrate this information into the traffic prediction model through knowledge representation learning methods. The Knowledge Graph can naturally describe the complex associations between different external factors, thereby providing richer background information for traffic prediction. Based on the above problems, this paper proposes a Knowledge - Driven Spatial - Temporal Graph Convolutional Network (KST - GCN), which specifically includes the following aspects: - **Constructing a knowledge graph**: Represent traffic information and multiple external factors as a heterogeneous semantic network, that is, a knowledge graph. - **Knowledge representation learning**: Adopt a knowledge representation learning method named KR - EAR to capture the semantic relationships between traffic information and external factors. - **Knowledge Fusion Cell (KF - Cell)**: Design a knowledge fusion cell to combine the knowledge extracted from the knowledge graph with traffic features and pass it as input to the backbone network of the spatial - temporal graph convolutional network. In this way, KST - GCN can not only capture the spatio - temporal features of traffic flow, but also effectively integrate the information of external factors, thereby improving the accuracy of traffic prediction. The experimental results show that within different prediction time ranges, after adding KF - Cell, the prediction performance of various backbone models has been improved. In addition, the ablation experiment further verifies the effectiveness of semantic relationships in traffic prediction and the impact of dynamic and static external factors on prediction results.

KST-GCN: A Knowledge-Driven Spatial-Temporal Graph Convolutional Network for Traffic Forecasting

A Spatial-Temporal Gated Hypergraph Convolution Network for Traffic Prediction

T-GCN: A Temporal Graph ConvolutionalNetwork for Traffic Prediction

T-GCN: A Temporal Graph Convolutional Network for Traffic Prediction

Graph learning-based spatial-temporal graph convolutional neural networks for traffic forecasting

Local-Global Spatial-Temporal Graph Convolutional Network for Traffic Flow Forecasting

Spatial–Temporal-Correlation-Constrained Dynamic Graph Convolutional Network for Traffic Flow Forecasting

Dual Dynamic Spatial-Temporal Graph Convolution Network for Traffic Prediction

Probabilistic spatio-temporal graph convolutional network for traffic forecasting

Spatio-Temporal Graph Convolutional Networks: A Deep Learning Framework for Traffic Forecasting

DMGSTCN: Dynamic Multi-Graph Spatio-Temporal Convolution Network for Traffic Forecasting

Spatio-Temporal Joint Graph Convolutional Networks for Traffic Forecasting

Spatial-Temporal Interactive Dynamic Graph Convolution Network for Traffic Forecasting

Spatial–Temporal Dynamic Graph Convolutional Network With Interactive Learning for Traffic Forecasting

Lanczos Method for Spatio‐temporal Graph Convolutional Networks to Forecast Expressway Flow

Hierarchical Spatio–Temporal Graph Convolutional Networks and Transformer Network for Traffic Flow Forecasting

STN-GCN: Spatial and Temporal Normalization Graph Convolutional Neural Networks for Traffic Flow Forecasting

Spatial–Temporal Complex Graph Convolution Network for Traffic Flow Prediction

3D Graph Convolutional Networks with Temporal Graphs: A Spatial Information Free Framework For Traffic Forecasting

PKET-GCN: Prior Knowledge Enhanced Time-Varying Graph Convolution Network for Traffic Flow Prediction