Daily Arxiv

This is a page that curates AI-related papers published worldwide.
All content here is summarized using Google Gemini and operated on a non-profit basis.
Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.

M3-Net: A Cost-Effective Graph-Free MLP-Based Model for Traffic Prediction

Created by
  • Haebom

Author

Guangyin Jin, Sicong Lai, Xiaoshuai Hao, Mingtao Zhang, Jinlei Zhang

Outline

This paper proposes M3-Net, a cost-effective, graph-free multilayer perceptron (MLP)-based model for accurate traffic volume prediction, essential for the development of intelligent transportation systems. Existing deep learning-based traffic volume prediction models either rely on the complete transportation network structure or require complex model designs to capture complex spatiotemporal dependencies. M3-Net addresses these issues by utilizing time-series and spatiotemporal embeddings for efficient feature processing and introducing a novel MLP-Mixer architecture with a Mixture of Experts (MoE) mechanism. Experiments on various real-world datasets demonstrate the superior predictive performance and lightweight deployability of the proposed model.

Takeaways, Limitations

•
Takeaways:
◦
A lightweight MLP-based traffic volume prediction model that overcomes the limitations of graph-based models is presented.
◦
Utilizing time series and spatiotemporal embeddings for efficient feature processing.
◦
Performance improvement through the introduction of the MLP-Mixer architecture utilizing the MoE mechanism.
◦
Validated for excellent predictive performance and lightweight deployability on various real-world datasets.
•
Limitations:
◦
Further research is needed on the generalization performance of the proposed model.
◦
Applicability assessment for transportation networks of various sizes is needed.
◦
Analysis of the computational costs of the MoE mechanism is needed.
👍
Made with Slashpage