Pets: General Pattern Assisted Architecture For Time Series Analysis

2025-04-19

AI Summary

Real-world time series often exhibit intertwined periodic fluctuations at multiple granularities (e.g., hourly, daily, monthly), which conventional decomposition methods struggle to disentangle without prior knowledge, thereby limiting analytical performance. To address this, we propose an adaptive frequency-band quantization method grounded in time-frequency energy distribution—establishing, for the first time, a time-frequency energy-driven paradigm for multi-periodic pattern disentanglement. We further design a plug-and-play Fluctuation Pattern Auxiliary (FPA) module and a context-guided Mixture-of-Predictors (MoP) mechanism, enabling cross-task and cross-architecture generalization. Our approach integrates implicit multi-scale modeling, hierarchical fluctuation reconstruction, and gated mixture prediction. It achieves state-of-the-art performance across four fundamental tasks—forecasting, imputation, anomaly detection, and classification—demonstrating strong generalizability, robustness, and seamless compatibility with arbitrary backbone architectures.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Computer Vision: Multi-modal Vision
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

Time series analysis has found widespread applications in areas such as weather forecasting, anomaly detection, and healthcare. However, real-world sequential data often exhibit a superimposed state of various fluctuation patterns, including hourly, daily, and monthly frequencies. Traditional decomposition techniques struggle to effectively disentangle these multiple fluctuation patterns from the seasonal components, making time series analysis challenging. Surpassing the existing multi-period decoupling paradigms, this paper introduces a novel perspective based on energy distribution within the temporal-spectrum space. By adaptively quantifying observed sequences into continuous frequency band intervals, the proposed approach reconstructs fluctuation patterns across diverse periods without relying on domain-specific prior knowledge. Building upon this innovative strategy, we propose Pets, an enhanced architecture that is adaptable to arbitrary model structures. Pets integrates a Fluctuation Pattern Assisted (FPA) module and a Context-Guided Mixture of Predictors (MoP). The FPA module facilitates information fusion among diverse fluctuation patterns by capturing their dependencies and progressively modeling these patterns as latent representations at each layer. Meanwhile, the MoP module leverages these compound pattern representations to guide and regulate the reconstruction of distinct fluctuations hierarchically. Pets achieves state-of-the-art performance across various tasks, including forecasting, imputation, anomaly detection, and classification, while demonstrating strong generalization and robustness.

Problem

• Disentangle multiple fluctuation patterns in time series data
• Adaptively quantify sequences into frequency band intervals
• Enhance time series analysis without domain-specific prior knowledge

Innovation

• Energy distribution in temporal-spectrum space
• Fluctuation Pattern Assisted module integration
• Context-Guided Mixture of Predictors