Research on Traffic Flow Forecasting Using Multi-Scale Spatial-Temporal Modeling and Spectrum Enhancement.

Saved in:

Bibliographic Details
Title:	Research on Traffic Flow Forecasting Using Multi-Scale Spatial-Temporal Modeling and Spectrum Enhancement.
Authors:	Sun, Baoyu¹ 13897812832@163.com, Dai, Hong² dear_red9@163.com, Zhang, Jicheng¹ 839519835@qq.com
Source:	Engineering Letters. May2026, Vol. 34 Issue 5, p1649-1657. 9p.
Subjects:	Multiscale modeling, Spatiotemporal processes, Signal convolution, Attention, Traffic estimation, Intelligent transportation systems, Deep learning
Abstract:	With the rapid development of intelligent transportation systems and smart mobility, accurate traffic flow prediction plays a crucial role in traffic scheduling, congestion management, and travel planning. To address the strong temporal dependencies and complex spatial correlations inherent in traffic flow data, this paper proposes a Transformer based traffic flow prediction model termed Fourier Multi-Scale-Transformer (FMS-Transformer). At the feature embedding stage, a Spectrum Enhancement (SE) mechanism is introduced to integrate time-of-day features, day-of-week features, and node-adaptive embedding, thereby enhancing the representation of periodic and heterogeneous spatial-temporal features. Subsequently, spatial-temporal self-attention mechanisms are employed to model global dependencies along the temporal and spatial dimensions, respectively. Furthermore, a Multi-Scale Deep Separable Temporal Convolution (MS-DS Conv) is incorporated to capture short-term fluctuations and long-term trends using one dimensional temporal convolution kernels with different scales, while an attention fusion mechanism is adopted to adaptively integrate multi-scale temporal features. Experimental results on the PEMS08 dataset demonstrate that the proposed FMS-Transformer consistently outperforms baseline models in terms of root mean square error, mean absolute error, and mean absolute percentage error, thereby validating its effectiveness and superiority in traffic flow prediction tasks. [ABSTRACT FROM AUTHOR]
	Copyright of Engineering Letters is the property of International Association of Engineers (IAENG) and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database:	Engineering Source

Description
Abstract:	With the rapid development of intelligent transportation systems and smart mobility, accurate traffic flow prediction plays a crucial role in traffic scheduling, congestion management, and travel planning. To address the strong temporal dependencies and complex spatial correlations inherent in traffic flow data, this paper proposes a Transformer based traffic flow prediction model termed Fourier Multi-Scale-Transformer (FMS-Transformer). At the feature embedding stage, a Spectrum Enhancement (SE) mechanism is introduced to integrate time-of-day features, day-of-week features, and node-adaptive embedding, thereby enhancing the representation of periodic and heterogeneous spatial-temporal features. Subsequently, spatial-temporal self-attention mechanisms are employed to model global dependencies along the temporal and spatial dimensions, respectively. Furthermore, a Multi-Scale Deep Separable Temporal Convolution (MS-DS Conv) is incorporated to capture short-term fluctuations and long-term trends using one dimensional temporal convolution kernels with different scales, while an attention fusion mechanism is adopted to adaptively integrate multi-scale temporal features. Experimental results on the PEMS08 dataset demonstrate that the proposed FMS-Transformer consistently outperforms baseline models in terms of root mean square error, mean absolute error, and mean absolute percentage error, thereby validating its effectiveness and superiority in traffic flow prediction tasks. [ABSTRACT FROM AUTHOR]
ISSN:	1816093X