Dynamic spatio-temporal graph network with adaptive propagation mechanism for multivariate time series forecasting

Zhuolin Li, Jie Yu, Gaowei Zhang, Lingyu Xu

Abstract: Highlights•Modeling dynamic dependencies among variables with proposed graph matrix estimation.•Adaptive guided propagation can change the propagation and aggregation process.•Multiple losses are designed to jointly optimize the network.•Experiment results demonstrate the effectiveness of proposed method.