Multivariate Forecasting

Multivariate forecasting aims to predict multiple interconnected time series, crucial for applications ranging from energy management to financial modeling. Current research emphasizes improving accuracy and efficiency through advanced architectures like transformers, graph neural networks, and hybrid models combining techniques such as Variational Mode Decomposition with linear models or incorporating Independent Component Analysis for enhanced feature representation. These advancements are driven by the need to handle complex dependencies between variables, noisy data, and distribution shifts, ultimately leading to more reliable and actionable predictions across diverse domains.

Papers