Spatial linear transformer and temporal convolution network for traffic flow prediction

Zhibo Xing,Mingxia Huang,Wentao Li,Dan Peng

DOI: https://doi.org/10.1038/s41598-024-54114-9

IF: 4.6

2024-02-20

Scientific Reports

Abstract:Accurately obtaining accurate information about the future traffic flow of all roads in the transportation network is essential for traffic management and control applications. In order to address the challenges of acquiring dynamic global spatial correlations between transportation links and modeling time dependencies in multi-step prediction, we propose a spatial linear transformer and temporal convolution network (SLTTCN). The model is using spatial linear transformers to aggregate the spatial information of the traffic flow, and bidirectional temporal convolution network to capture the temporal dependency of the traffic flow. The spatial linear transformer effectively reduces the complexity of data calculation and storage while capturing spatial dependence, and the time convolutional network with bidirectional and gate fusion mechanisms avoids the problems of gradient vanishing and high computational cost caused by long time intervals during model training. We conducted extensive experiments using two publicly available large-scale traffic data sets and compared SLTTCN with other baselines. Numerical results show that SLTTCN achieves the best predictive performance in various error measurements. We also performed attention visualization analysis on the spatial linear transformer, verifying its effectiveness in capturing dynamic global spatial dependency.

multidisciplinary sciences

What problem does this paper attempt to address?

This paper aims to solve several key problems in traffic flow prediction:

1. **Modeling of dynamic global spatial correlations**: In the traffic network, the spatial correlations between different road sections are dynamically changing. Traditional methods based on pre - defined graph structures may not accurately reflect this dynamic relationship. The paper proposes a Spatial Linear Transformer (SLT), which dynamically captures spatial dependencies through real - time traffic states, connectivity between nodes, and traffic flow directions.

2. **Modeling of time - dependent relationships**: The time - dependent relationships of traffic flow involve long - and short - term time intervals. Traditional time - series models such as RNN and its variants have problems such as high computational complexity and vanishing gradients when dealing with long - term dependencies. The paper introduces a Temporal Convolution Network (TCN) and uses a Gate Fusion Mechanism to capture complex time - dependent relationships from the past to the future and from the future to the past.

3. **Computational efficiency and resource requirements**: In order to improve the computational efficiency of the model and reduce resource requirements, the paper optimizes the traditional self - attention mechanism and proposes a Linear Self - Attention mechanism, thereby reducing memory and computational complexity while maintaining model performance.

In summary, by proposing a new framework - Spatial Linear Transformer with Temporal Convolution Network (SLTTCN), this paper aims to more effectively model the spatio - temporal dependencies of traffic flow and has verified its superior prediction performance on multiple datasets.