TimeFilter: Patch-Specific Spatial-Temporal Graph Filtration for Time Series Forecasting

2025-01-22

AI Summary

Existing channel-independent (CI) methods for multivariate time series forecasting neglect inter-variable correlations, while channel-dependent (CD) approaches often introduce spurious dependencies. To address these limitations, this paper proposes a dynamic graph-based fine-grained modeling framework. Its core innovation is the first-ever patch-specific spatiotemporal graph filtering mechanism, enabling time-varying and locally adaptive correlation selection—thereby overcoming the coarse-grained constraints of conventional channel clustering. The framework further integrates graph neural networks, sliding-block modeling, learnable spatiotemporal graph filters, and channel-aware adjacency matrix optimization. Extensive experiments across 13 real-world datasets from diverse domains demonstrate consistent superiority over CI, CD, and channel-correlated (CC) baselines, achieving state-of-the-art performance. The implementation is publicly available.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Mining of Spatial, Temporal or Spatio-Temporal Data
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Web Mining and Content Analysis: Models for Web evolution
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Current time series forecasting methods can be broadly classified into two categories: Channel Independent (CI) and Channel Dependent (CD) strategies, both aiming to capture the complex dependencies within time series data. However, the CI strategy fails to exploit highly correlated covariate information, while the CD strategy integrates all dependencies, including irrelevant or noisy ones, thus compromising generalization. To mitigate these issues, recent works have introduced the Channel Clustering (CC) strategy by grouping channels with similar characteristics and applying different modeling techniques to each cluster. However, coarse-grained clustering cannot flexibly capture complex, time-varying interactions. Addressing the above challenges, we propose TimeFilter, a graph-based framework for adaptive and fine-grained dependency modeling. Specifically, after constructing the graph with the input sequence, TimeFilter filters out irrelevant correlations and preserves the most critical ones through patch-specific filtering. Extensive experiments on 13 real-world datasets from various application domains demonstrate the state-of-the-art performance of TimeFilter. The code is available at https://github.com/TROUBADOUR000/TimeFilter.

Problem

• Time Series Prediction
• Channel Independence
• Channel Dependency

Innovation

• TimeFilter
• spatiotemporal graph filtering
• performance enhancement