DGCformer: Deep Graph Clustering Transformer for Multivariate Time Series Forecasting

TL;DR

The paper addresses forecasting accuracy for multivariate time series by balancing channel dependence and independence through channel semi-dependence. It introduces DGCformer, which clusters variables using a GRU-based Variational Autoencoder followed by $k$-means to form groups and then applies group-wise CD self-attention within a PatchTST-based transformer, while treating inter-group interactions with CI. The training objective combines reconstruction, KL divergence, and prediction losses via $\mathcal{L}_{total}=\mathcal{L}_{REC}+\mathcal{L}_{KL}+\mathcal{L}_{PRED}$. Empirically, it achieves superior results on eight datasets, demonstrating effective variable grouping and cross-group interaction, with code to be released.$

Abstract

Multivariate time series forecasting tasks are usually conducted in a channel-dependent (CD) way since it can incorporate more variable-relevant information. However, it may also involve a lot of irrelevant variables, and this even leads to worse performance than the channel-independent (CI) strategy. This paper combines the strengths of both strategies and proposes the Deep Graph Clustering Transformer (DGCformer) for multivariate time series forecasting. Specifically, it first groups these relevant variables by a graph convolutional network integrated with an autoencoder, and a former-latter masked self-attention mechanism is then considered with the CD strategy being applied to each group of variables while the CI one for different groups. Extensive experimental results on eight datasets demonstrate the superiority of our method against state-of-the-art models, and our code will be publicly available upon acceptance.