🤖 AI Summary
This work addresses the challenge that existing Mamba architectures struggle to effectively model the heterogeneous relationships between trend and seasonal components in non-stationary time series. To this end, it pioneers the integration of seasonal-trend decomposition with the Mamba framework, devising distinct modeling strategies tailored to the intrinsic characteristics of each component: a highly expressive variable-direction Mamba encoder captures the dynamic, high-dimensional interactions of seasonal patterns, while a lightweight MLP models the low-dimensional, long-term equilibrium relationships inherent in the trend component. Evaluated across multiple benchmark datasets, the proposed method significantly outperforms current Mamba variants and state-of-the-art decomposition-based models, achieving new state-of-the-art performance and demonstrating a precise alignment between model architecture and the statistical properties of time series data.
📝 Abstract
State Space Models (SSMs), particularly Mamba, have shown potential in long-term time series forecasting. However, existing Mamba-based architectures often struggle with datasets characterized by non-stationary patterns. A key observation from time series theory is that the statistical nature of inter-variable relationships differs fundamentally between the trend and seasonal components of a decomposed series. Trend relationships are often driven by a few common stochastic factors or long-run equilibria, suggesting that they reside on a lower-dimensional manifold. In contrast, seasonal relationships involve dynamic, high-dimensional interactions like phase shifts and amplitude co-movements, requiring more expressive modeling. In this paper, we propose DMamba, a novel forecasting model that explicitly aligns architectural complexity with this component-specific characteristic. DMamba employs seasonal-trend decomposition and processes the components with specialized, differentially complex modules: a variable-direction Mamba encoder captures the rich, cross-variable dynamics within the seasonal component, while a simple Multi-Layer Perceptron (MLP) suffices to learn from the lower-dimensional inter-variable relationships in the trend component. Extensive experiments on diverse datasets demonstrate that DMamba sets a new state-of-the-art (SOTA), consistently outperforming both recent Mamba-based architectures and leading decomposition-based models.