Multivariate time series (MTS) data representation learning has received great attention in recent years due to its importance for downstream tasks. MTS has two important properties: the temporal and spatial dependencies. Traditionally, owing to highly nonlinear power, deep -learning-based methods have been widely applied for addressing the former, but few works focus on making full of the spatial dependencies. Although pioneers started to exploit a graph neural network to capture the spatial dependencies, they still have limitations. The proposed method in this article learns and captures the hierarchical correlations between sensors, and meanwhile, sequential graphs are learned to represent the dynamic property within MTS data. In this way, the spatialtemporal dependencies within MTS data can be better leveraged. With significant improvements on multiple MTS tasks compared to state-of-the-art algorithms, our method is ready to learn decent representations from MTS signals in differen...

Representation learning is vital for the performance of multivariate time series (MTS)-related tasks. Given high-dimensional MTS data, researchers generally rely on deep learning models to learn representative features. Among them, the methods that can capture the spatial–temporal dependencies within MTS data generally achieve better performance. However, they ignored hierarchical relations and the dynamic property within MTS data, hindering their performance. To address these problems, we propose a hierarchical correlation pooling boosted graph neural network for MTS data representation learning. First, we propose a novel correlation pooling scheme to learn and capture hierarchical correlations between sensors. Meanwhile, a new assignment matrix is designed to ensure the effective learning of hierarchical correlations by adaptively combining both sensor features and correlations. Second, we learn sequential graphs to represent the dynamic property within MTS data, so that this propert...