STAMImputer: Spatio-Temporal Attention MoE for Traffic Data Imputation
By: Yiming Wang , Hao Peng , Senzhang Wang and more
Potential Business Impact:
Fixes missing traffic data for smarter roads.
Traffic data imputation is fundamentally important to support various applications in intelligent transportation systems such as traffic flow prediction. However, existing time-to-space sequential methods often fail to effectively extract features in block-wise missing data scenarios. Meanwhile, the static graph structure for spatial feature propagation significantly constrains the models flexibility in handling the distribution shift issue for the nonstationary traffic data. To address these issues, this paper proposes a SpatioTemporal Attention Mixture of experts network named STAMImputer for traffic data imputation. Specifically, we introduce a Mixture of Experts (MoE) framework to capture latent spatio-temporal features and their influence weights, effectively imputing block missing. A novel Low-rank guided Sampling Graph ATtention (LrSGAT) mechanism is designed to dynamically balance the local and global correlations across road networks. The sampled attention vectors are utilized to generate dynamic graphs that capture real-time spatial correlations. Extensive experiments are conducted on four traffic datasets for evaluation. The result shows STAMImputer achieves significantly performance improvement compared with existing SOTA approaches. Our codes are available at https://github.com/RingBDStack/STAMImupter.
Similar Papers
STGAtt: A Spatial-Temporal Unified Graph Attention Network for Traffic Flow Forecasting
Machine Learning (CS)
Predicts traffic jams before they happen.
MSTIM: A MindSpore-Based Model for Traffic Flow Prediction
Machine Learning (CS)
Predicts traffic jams more accurately.
Lightweight Spatio-Temporal Attention Network with Graph Embedding and Rotational Position Encoding for Traffic Forecasting
Artificial Intelligence
Predicts traffic jams much better.