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

M3-Net represents a breakthrough in traffic prediction by addressing the challenges posed by traditional deep learning methods that depend on intricate spatio-temporal graph structures. The model employs a novel MLP-Mixer architecture with a mixture of experts (MoE) mechanism, allowing for efficient feature processing through time series and spatio-temporal embeddings. Extensive experiments on real datasets have demonstrated M3-Net's superiority in prediction performance and its lightweight deployment capabilities. This innovation is crucial for the development of intelligent transportation systems, which require accurate traffic forecasting to enhance traffic management and operational efficiency.