Time-aware personalized graph convolutional network for multivariate time series forecasting

Zhuolin Li, Ziheng Gao, Xiaolin Zhang, Gaowei Zhang, Lingyu Xu

Abstract: Highlights•Proposed graph learning method models spatial dependence under heterogeneous information.•Variables are impacted by self-evolutionary pattern and external variables.•Proposed graph convolution unifies the effects of self and external variables into a framework.•Proposed framework attains state-of-the-art results on benchmark datasets.