Flow prediction via adaptive dynamic graph with spatio-temporal correlations

Hui Zhang, Kun Ding, Jietao Xie, Weidong Xiao, Yuxiang Xie

Abstract: Highlights•Capture spatial correlations in graphs as well as short-term and long-term time dependencies.•Propose node-adaptive learning that alleviates computational burden and enhances efficiency.•Devise graph adaptive generation that improves the ability to identify spatial relations.•Conduct multi-step prediction via identifying node-specific spatio-temporal patterns.