TKSTAGNet: A Top-K Spatio-Temporal Attention Gating Network for air pollution prediction

Hongliang Zhou, Yingchi Mao, Xiang Li, Yi Rong, Ling Chen, Changkui Yin

Abstract: Highlights•We propose a novel Transformer architecture for air pollution prediction.•We develop a gating fusion module to adaptively fuse external factors.•Our proposed model outperforms advanced baselines in prediction accuracy.