Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

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

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

TL;DR

Traffic forecasting in large urban networks is challenged by reliance on graph topologies and costly ST models. M3-Net presents a graph-free MLP architecture that combines a Spatial MLP with an adaptive grouping mechanism and a MoE-enhanced Channel MLP, guided by rich spatio-temporal embeddings, to model multi-scale ST dynamics directly from raw sequences. The approach achieves state-of-the-art performance on multiple real datasets while maintaining light-weight deployment, as shown by comprehensive ablations, visualization of learned groupings, and cost-efficiency analyses. This work enables scalable, efficient traffic prediction suitable for deployment on resource-constrained ITS platforms without explicit topology constraints.

Abstract

Achieving accurate traffic prediction is a fundamental but crucial task in the development of current intelligent transportation systems.Most of the mainstream methods that have made breakthroughs in traffic prediction rely on spatio-temporal graph neural networks, spatio-temporal attention mechanisms, etc. The main challenges of the existing deep learning approaches are that they either depend on a complete traffic network structure or require intricate model designs to capture complex spatio-temporal dependencies. These limitations pose significant challenges for the efficient deployment and operation of deep learning models on large-scale datasets. To address these challenges, we propose a cost-effective graph-free Multilayer Perceptron (MLP) based model M3-Net for traffic prediction. Our proposed model not only employs time series and spatio-temporal embeddings for efficient feature processing but also first introduces a novel MLP-Mixer architecture with a mixture of experts (MoE) mechanism. Extensive experiments conducted on multiple real datasets demonstrate the superiority of the proposed model in terms of prediction performance and lightweight deployment.Our code is available at https://github.com/jinguangyin/M3_NET

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

TL;DR

Traffic forecasting in large urban networks is challenged by reliance on graph topologies and costly ST models. M3-Net presents a graph-free MLP architecture that combines a Spatial MLP with an adaptive grouping mechanism and a MoE-enhanced Channel MLP, guided by rich spatio-temporal embeddings, to model multi-scale ST dynamics directly from raw sequences. The approach achieves state-of-the-art performance on multiple real datasets while maintaining light-weight deployment, as shown by comprehensive ablations, visualization of learned groupings, and cost-efficiency analyses. This work enables scalable, efficient traffic prediction suitable for deployment on resource-constrained ITS platforms without explicit topology constraints.

Abstract

Achieving accurate traffic prediction is a fundamental but crucial task in the development of current intelligent transportation systems.Most of the mainstream methods that have made breakthroughs in traffic prediction rely on spatio-temporal graph neural networks, spatio-temporal attention mechanisms, etc. The main challenges of the existing deep learning approaches are that they either depend on a complete traffic network structure or require intricate model designs to capture complex spatio-temporal dependencies. These limitations pose significant challenges for the efficient deployment and operation of deep learning models on large-scale datasets. To address these challenges, we propose a cost-effective graph-free Multilayer Perceptron (MLP) based model M3-Net for traffic prediction. Our proposed model not only employs time series and spatio-temporal embeddings for efficient feature processing but also first introduces a novel MLP-Mixer architecture with a mixture of experts (MoE) mechanism. Extensive experiments conducted on multiple real datasets demonstrate the superiority of the proposed model in terms of prediction performance and lightweight deployment.Our code is available at https://github.com/jinguangyin/M3_NET

Paper Structure

This paper contains 13 sections, 4 figures, 3 tables.

Table of Contents

  1. INTRODUCTION
  2. METHODOLOGY
  3. Embedding Layer
  4. MLP-Mixer with MoE
  5. Regression Layer
  6. EXPERIMENTS
  7. Experimental Setup
  8. Performance Evaluation
  9. Ablation Study
  10. Spatial Homogeneity Visualization
  11. Cost-Effectiveness Analysis
  12. CONCLUSION
  13. GenAI Usage Disclosure

Figures (4)

  • Figure 1: Spatio-Temporal Characteristics of Traffic Flow.
  • Figure 2: The framework of M3-Net.
  • Figure 3: Adaptive grouping matrices visualization on PEMS04 and PEMS08.
  • Figure 4: Visualization of cost-effectiveness of deployment on PEMS08 dataset.