Spatial temporal network improves forecasting by separating trend and seasonality
DST-Net: Decomposed Spatio-Temporal Network for Time Series Forecasting
Machine LearningArtificial Intelligence
Summary
Forecasting future data points in time series is tricky because trends and seasonal patterns act differently. The authors propose a new model that treats the trend and seasonal parts separately to handle their unique behaviors better. For trends, the model uses a graph-based method to capture relationships across different channels. For seasonal parts, it uses a focused attention method to avoid distractions from irrelevant data. Their approach shows improved accuracy on multiple real-world datasets.
What this means in practice
- •For data scientists in finance: Improve forecasting models for multivariate financial data by separately modeling trend and seasonal components to capture complex dependencies.
- •For energy demand planners: Enhance energy consumption forecasts by effectively capturing global trends and local seasonal variations in multi-channel time series data.
Authors
Yaohua Zha, Hai Wu, Shu-Tao Xia
Abstract
Series decomposition has emerged as a widely acknowledged paradigm in deep time series forecasting models. However, existing approaches predominantly employ a homogeneous modeling strategy for all subsequences, thereby overlooking the distinct inductive biases inherent in different components. We identify that the trend component exhibits dominant spatial coupling across channels but carries significant temporal redundancy, whereas the seasonal component is characterized by local temporal evolution but remains susceptible to attention distraction. To address this, we propose a novel disentangled architecture that tailors specific modeling strategies to align with the intrinsic properties of each component. For the trend branch, we design a Latent Graph Block, employing latent graph modeling to capture global topological dependencies and system-wide co-movements. Conversely, for the seasonal branch, we introduce a Gaussian Multi-scale Attention mechanism to mitigate attention distraction. By injecting learnable Gaussian priors, we impose structural constraints that compel the model to focus on multi-granular local contexts, effectively filtering out global noise and accurately capturing local evolutionary phases. Extensive experiments on multiple real-world benchmarks demonstrate the effectiveness of our proposed model.