LLM-Mixer: Multiscale Mixing in LLMs for Time Series Forecasting

2024-10-15

arXiv.org

AI Summary

Time-series forecasting remains challenging under complex multi-scale dynamics. This paper proposes a novel paradigm that synergistically integrates multi-scale time-series decomposition (e.g., EMD, wavelet) with frozen large language models (LLMs) such as Llama and Qwen. Leveraging time-aware textual prompts, the approach guides LLMs to jointly reason about trends and fluctuations without fine-tuning—marking the first seamless integration of pre-trained LLMs into decomposition frameworks. The method fuses multi-resolution features while preserving interpretability and generalizability. Extensive experiments on both multivariate and univariate benchmark datasets demonstrate consistent superiority over recent state-of-the-art methods across short-, medium-, and long-term forecasting tasks. Results validate the feasibility and effectiveness of frozen LLMs as universal time-series reasoning engines.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Time series forecasting remains a challenging task, particularly in the context of complex multiscale temporal patterns. This study presents LLM-Mixer, a framework that improves forecasting accuracy through the combination of multiscale time-series decomposition with pre-trained LLMs (Large Language Models). LLM-Mixer captures both short-term fluctuations and long-term trends by decomposing the data into multiple temporal resolutions and processing them with a frozen LLM, guided by a textual prompt specifically designed for time-series data. Extensive experiments conducted on multivariate and univariate datasets demonstrate that LLM-Mixer achieves competitive performance, outperforming recent state-of-the-art models across various forecasting horizons. This work highlights the potential of combining multiscale analysis and LLMs for effective and scalable time-series forecasting.

Problem

• Improves time series forecasting accuracy with multiscale decomposition
• Combines LLMs and multiscale analysis for complex temporal patterns
• Addresses short-term fluctuations and long-term trends in forecasting

Innovation

• Multiscale time-series decomposition with LLMs
• Frozen LLM processing guided by textual prompts
• Combining multiscale analysis for scalable forecasting