BLAST: Balanced Sampling Time Series Corpus for Universal Forecasting Models

2025-05-23

AI Summary

Existing large-scale time-series datasets suffer from distributional skewness and uneven pattern coverage, severely limiting the generalization capability of pretrained models. To address this, we propose a balanced sampling strategy for constructing a pretraining corpus tailored to generic time-series forecasting: we introduce implicit clustering via statistical feature-based grid partitioning, coupled with Grid Mixup—a novel pattern-aware sampling and augmentation technique—to ensure equitable coverage of temporal patterns and enhance data diversity. The resulting corpus significantly improves both the diversity and representativeness of time-series patterns. Building upon this, we design a lightweight, efficient large-scale pretraining framework that achieves state-of-the-art (SOTA) performance on zero-shot forecasting tasks—using only ∼1/3 of the computational resources and training tokens required by prior approaches—while substantially enhancing cross-domain generalization and training efficiency.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Search and Optimization: Sampling/Simulation-based Search
• Data Mining & Knowledge Management: Scalability, Parallel & Distributed Systems

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• User Modeling, Personalization and Recommendation: Practical large-scale studies of user experience
• Search and Retrieval-Augmented AI: Vertical and domain-specific search

Abstract

The advent of universal time series forecasting models has revolutionized zero-shot forecasting across diverse domains, yet the critical role of data diversity in training these models remains underexplored. Existing large-scale time series datasets often suffer from inherent biases and imbalanced distributions, leading to suboptimal model performance and generalization. To address this gap, we introduce BLAST, a novel pre-training corpus designed to enhance data diversity through a balanced sampling strategy. First, BLAST incorporates 321 billion observations from publicly available datasets and employs a comprehensive suite of statistical metrics to characterize time series patterns. Then, to facilitate pattern-oriented sampling, the data is implicitly clustered using grid-based partitioning. Furthermore, by integrating grid sampling and grid mixup techniques, BLAST ensures a balanced and representative coverage of diverse patterns. Experimental results demonstrate that models pre-trained on BLAST achieve state-of-the-art performance with a fraction of the computational resources and training tokens required by existing methods. Our findings highlight the pivotal role of data diversity in improving both training efficiency and model performance for the universal forecasting task.

Problem

• Addressing data diversity gaps in universal time series forecasting models
• Mitigating biases and imbalances in large-scale time series datasets
• Enhancing model performance through balanced pattern-oriented sampling strategies

Innovation

• Balanced sampling strategy enhances data diversity
• Grid-based partitioning for pattern-oriented sampling
• Grid sampling and mixup ensure representative coverage