Temporal Patch Shuffle (TPS): Leveraging Patch-Level Shuffling to Boost Generalization and Robustness in Time Series Forecasting

2026-04-10

AI Summary

This work addresses the challenge that existing time series data augmentation methods often degrade forecasting performance by disrupting temporal consistency. To overcome this limitation, the authors propose TPS, a model-agnostic augmentation strategy that introduces a novel segment-level shuffling mechanism. Specifically, overlapping temporal segments are extracted, and those exhibiting high variance—identified via variance-based ranking—are selectively shuffled. During reconstruction, overlapping regions are smoothed by averaging, thereby preserving local temporal structures while enhancing data diversity. Evaluated across nine long-term and four short-term forecasting benchmark datasets in conjunction with five state-of-the-art forecasting models, TPS consistently yields significant performance improvements, demonstrating its effectiveness and robustness.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Planning, Routing, and Scheduling: Temporal Planning
• Data Mining & Knowledge Management: Mining of Spatial, Temporal or Spatio-Temporal Data

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Security and Privacy: Data transparency and provenance

Abstract

Data augmentation is a crucial technique for improving model generalization and robustness, particularly in deep learning models where training data is limited. Although many augmentation methods have been developed for time series classification, most are not directly applicable to time series forecasting due to the need to preserve temporal coherence. In this work, we propose Temporal Patch Shuffle (TPS), a simple and model-agnostic data augmentation method for forecasting that extracts overlapping temporal patches, selectively shuffles a subset of patches using variance-based ordering as a conservative heuristic, and reconstructs the sequence by averaging overlapping regions. This design increases sample diversity while preserving forecast-consistent local temporal structure. We extensively evaluate TPS across nine long-term forecasting datasets using five recent model families (TSMixer, DLinear, PatchTST, TiDE, and LightTS), and across four short-term forecasting datasets using PatchTST, observing consistent performance improvements. Comprehensive ablation studies further demonstrate the effectiveness, robustness, and design rationale of the proposed method.

Problem

• time series forecasting
• data augmentation
• temporal coherence
• generalization
• robustness

Innovation

• Temporal Patch Shuffle
• data augmentation
• time series forecasting
• patch shuffling
• temporal coherence