Related Experiment Videos
Enhanced short-term traffic flow forecasting using a meteorological spatio-temporal transformer with multi-head
S Revathi1, M P Paulraj2, R N Devendra Kumar3
1Department of Computer Science and Engineering, Sri Ramakrishna Institute of Technology, Pachapalayam, Coimbatore, Tamil Nadu, 641010, India. revathi.cse@sritcbe.ac.in.
Scientific Reports
|May 8, 2026
Summary
This study introduces an advanced traffic prediction model, the Meteorological Spatio-Temporal Transformer Network with Multi-Head Attention and Whale Optimization Algorithm (MSTT-MHA-WOA), for better urban traffic management. The novel approach improves short-term traffic flow forecasting accuracy in complex, real-world conditions.
Area of Science:
- Urban planning and transportation engineering
- Artificial intelligence and machine learning
- Environmental science and sustainability
Background:
- Urbanization and increased vehicle ownership lead to significant traffic congestion, impacting society, economy, and environment.
- Effective urban traffic management relies on accurate short-term traffic flow prediction.
- Traditional models often fail to capture the complex, non-linear dynamics inherent in traffic data.
Purpose of the Study:
- To develop an optimized deep learning model for enhanced short-term traffic flow prediction.
- To integrate non-linear meteorological features into traffic flow forecasting.
- To leverage autonomous hyperparameter tuning for improved model performance.
Main Methods:
- Proposed an optimized synthesis model: Meteorological Spatio-Temporal Transformer Network with Multi-Head Attention and Whale Optimization Algorithm (MSTT-MHA-WOA).
- Integrated non-linear meteorological data as input features.
- Utilized the Whale Optimization Algorithm (WOA) for automated hyperparameter optimization.
Main Results:
- The MSTT-MHA-WOA model demonstrated robust and consistent performance.
- Evaluation was conducted across six diverse, real-world traffic scenarios.
- The model's generalization capability under complex conditions was validated.
Conclusions:
- The proposed MSTT-MHA-WOA framework offers a significant advancement in short-term traffic flow prediction.
- Integration of meteorological data and WOA-based hyperparameter tuning enhances prediction accuracy and model robustness.
- The model shows strong potential for effective urban traffic management and congestion mitigation strategies.