CATS: Enhancing Multivariate Time Series Forecasting by Constructing Auxiliary Time Series as Exogenous Variables

TL;DR

The paper tackles the paradox that univariate models often outperform multivariate approaches in multivariate time series forecasting (MTSF). It introduces Constructing Auxiliary Time Series (CATS), which generates Auxiliary Time Series (ATS) from the Original Time Series (OTS) to capture inter-series dependencies as exogenous inputs, enabling a simple predictor to leverage multivariate information without changing its architecture. The framework rests on three ATS principles—continuity, sparsity, and variability—and employs multiple ATS constructors to represent diverse inter-series relationships, while preserving an OTS shortcut for stability. Empirical results across nine real-world datasets show CATS achieving state-of-the-art performance with a lightweight 2-layer MLP predictor and substantially lower complexity than Transformer-based multivariate models, highlighting its efficiency and transferability. The work suggests potential extensions to larger ATS families and cross-domain applications (e.g., NLP, audio) and discusses limitations related to a fixed ATS count.

Abstract

For Multivariate Time Series Forecasting (MTSF), recent deep learning applications show that univariate models frequently outperform multivariate ones. To address the difficiency in multivariate models, we introduce a method to Construct Auxiliary Time Series (CATS) that functions like a 2D temporal-contextual attention mechanism, which generates Auxiliary Time Series (ATS) from Original Time Series (OTS) to effectively represent and incorporate inter-series relationships for forecasting. Key principles of ATS - continuity, sparsity, and variability - are identified and implemented through different modules. Even with a basic 2-layer MLP as core predictor, CATS achieves state-of-the-art, significantly reducing complexity and parameters compared to previous multivariate models, marking it an efficient and transferable MTSF solution.

Figures (7)

• Figure 1: Overall Architecture of the CATS Approach
• Figure 2: The Basic CATS Solution of the Shifting Problem
• Figure 3: 2D Temporal-contextual Attention-like Mechanism
• Figure 4: Channel Attention Score Calculation of Channel Sparsity
• Figure 5: Adaptive Temporal Cutoff of Temporal Sparsity