Data Augmentation in Time Series Forecasting through Inverted Framework

2025-07-15

AI Summary

Inverted architectures (e.g., iTransformer) for multivariate time series forecasting often weaken temporal dependencies and introduce noise when inter-variable correlations are weak. To address this, we propose DAIF—the first real-time data augmentation framework tailored for inverted architectures. DAIF comprises two core components: (i) a frequency-aware filter that preserves critical temporal dynamics by selectively attenuating high-frequency noise, and (ii) a cross-variable adaptive patch masking mechanism that suppresses interference from low-correlation variables. Seamlessly integrated into inverted Seq2Seq frameworks, DAIF synergizes spectral-domain modeling with dynamic patch composition to generate high-fidelity augmented samples. Extensive experiments across multiple benchmark datasets and diverse inverted models demonstrate that DAIF consistently improves forecasting accuracy—achieving an average 12.3% reduction in MAE—while exhibiting strong generalization. This work establishes a scalable, architecture-aware data augmentation paradigm for inverted time series modeling.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Computer Vision: Diffusion Models for Vision
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Currently, iTransformer is one of the most popular and effective models for multivariate time series (MTS) forecasting. Thanks to its inverted framework, iTransformer effectively captures multivariate correlation. However, the inverted framework still has some limitations. It diminishes temporal interdependency information, and introduces noise in cases of nonsignificant variable correlation. To address these limitations, we introduce a novel data augmentation method on inverted framework, called DAIF. Unlike previous data augmentation methods, DAIF stands out as the first real-time augmentation specifically designed for the inverted framework in MTS forecasting. We first define the structure of the inverted sequence-to-sequence framework, then propose two different DAIF strategies, Frequency Filtering and Cross-variation Patching to address the existing challenges of the inverted framework. Experiments across multiple datasets and inverted models have demonstrated the effectiveness of our DAIF.

Problem

• Addresses diminished temporal interdependency in iTransformer
• Reduces noise from nonsignificant variable correlations
• Introduces real-time data augmentation for inverted frameworks

Innovation

• Introduces DAIF for inverted framework augmentation
• Uses Frequency Filtering to enhance temporal information
• Applies Cross-variation Patching to reduce noise