When multi-view meets multi-level: A novel spatio-temporal transformer for traffic prediction

Jiaqi Lin, Qianqian Ren, Xingfeng Lv, Hui Xu, Yong Liu

Abstract: Highlights•Proposing an augmented learning paradigm for long and short-term traffic prediction.•Enhancing spatial dependency capture through multiple views and context broadcasting.•Experimental results show a 4.32% improvement compared to SOAT baselines.