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.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is that in multivariate time - series forecasting (MTSF), existing multivariate models often perform worse than univariate models. The author observes that although univariate models ignore the relationships between different sequences, they can still outperform complex multivariate models in many cases. This may be because: 1. The benchmark datasets in the real world may have inherently weak inter - sequence relationships. 2. Previous multivariate models may lack an effective structure to capture the inter - relationships between sequences, leading to over - fitting or fitting of wrong patterns. To solve these problems, the paper introduces a new method - Constructing Auxiliary Time - Series (CATS). CATS effectively represents and combines the relationships between sequences by generating Auxiliary Time - Series (ATS) as exogenous variables, thereby improving the forecasting performance. This method not only improves the accuracy of multivariate time - series forecasting but also significantly reduces the complexity of the model and the number of parameters, making CATS an efficient and transferable MTSF solution. ### Key Points: - **Problem Background**: Multivariate time - series forecasting is very important in fields such as finance, health, and transportation, but the challenge lies in the need to understand and model the interdependencies between multiple sequences. - **Existing Problems**: Existing multivariate models often perform poorly when dealing with these interdependencies, even worse than simple univariate models. - **Solution**: CATS captures and represents the interdependencies between the original time - series (OTS) by constructing Auxiliary Time - Series (ATS), thereby enhancing the capabilities of the forecasting model. - **Core Mechanism**: CATS uses three key principles - continuity, sparsity, and variability, and implements these principles through different modules to effectively extract and represent multivariate relationships. ### Specific Contributions: 1. **Proposing the CATS Method**: Based on the multivariate original time - series, construct auxiliary time - series, endowing the time - series predictor with the ability to efficiently capture the relationships between sequences, similar to the 2D attention mechanism. 2. **Identifying the Key Principles of ATS**: Continuity, sparsity, and variability, and designing specific modules to enforce these principles. 3. **Proposing Multiple ATS Constructors**: Provide adaptability to general MTSF datasets, demonstrating the flexibility of CATS in handling diverse MTSF challenges. ### Experimental Results: - **Main Experimental Results**: CATS performs excellently on multiple commonly - used MTSF datasets, with an average MSE better than existing MTSF models and ranking close to the first. - **Extended Experimental Results**: In more complex multivariate - relationship scenarios, such as the MultiX dataset, CATS also performs excellently, proving its effectiveness in handling more sequences. In conclusion, this paper successfully solves the problem of poor performance of existing models in multivariate time - series forecasting by introducing the CATS method, providing new research directions and technical means for this field.

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

Foreformer: an Enhanced Transformer-Based Framework for Multivariate Time Series Forecasting

TFEformer: Temporal Feature Enhanced Transformer for Multivariate Time Series Forecasting

Generalizable Memory-driven Transformer for Multivariate Long Sequence Time-series Forecasting

TCLN: A Transformer-based Conv-LSTM Network for Multivariate Time Series Forecasting

Enhancing Multivariate Time Series Forecasting with Mutual Information-driven Cross-Variable and Temporal Modeling

TwinS: Revisiting Non-Stationarity in Multivariate Time Series Forecasting

AFSTGCN: Prediction for multivariate time series using an adaptive fused spatial-temporal graph convolutional network

TriD-MAE: A Generic Pre-trained Model for Multivariate Time Series with Missing Values

Temporal Self-Attention-based Conv-LSTM Network for Multivariate Time Series Prediction

Are Self-Attentions Effective for Time Series Forecasting?

Cluster-aware attentive convolutional recurrent network for multivariate time-series forecasting

Expressing Multivariate Time Series as Graphs with Time Series Attention Transformer

Spatial-Temporal Convolutional Transformer Network for Multivariate Time Series Forecasting

Learning Structured Components: Towards Modular and Interpretable Multivariate Time Series Forecasting

Parallel Extraction of Long-term Trends and Short-term Fluctuation Framework for Multivariate Time Series Forecasting

Disentangling Structured Components: Towards Adaptive, Interpretable and Scalable Time Series Forecasting

TV-Net: Temporal-Variable feature harmonizing Network for multivariate time series classification and interpretation

A lightweight multi-layer perceptron for efficient multivariate time series forecasting

UniTST: Effectively Modeling Inter-Series and Intra-Series Dependencies for Multivariate Time Series Forecasting

Multivariate Time Series Forecasting with Causal-Temporal Attention Network.