TimeCMA: Towards LLM-Empowered Multivariate Time Series Forecasting via Cross-Modality Alignment

TL;DR

TimeCMA introduces a cross-modality alignment framework combining disentangled time series embeddings with robust prompt embeddings, significantly improving multivariate time series forecasting performance and efficiency.

Contribution

It presents a novel dual-branch encoding approach that aligns disentangled and entangled embeddings using LLMs, enhancing forecasting accuracy and reducing computational costs.

Findings

Outperforms state-of-the-art methods on eight datasets

Effectively balances disentangled and robust embeddings

Reduces inference time with last token embedding strategy

Abstract

Multivariate time series forecasting (MTSF) aims to learn temporal dynamics among variables to forecast future time series. Existing statistical and deep learning-based methods suffer from limited learnable parameters and small-scale training data. Recently, large language models (LLMs) combining time series with textual prompts have achieved promising performance in MTSF. However, we discovered that current LLM-based solutions fall short in learning disentangled embeddings. We introduce TimeCMA, an intuitive yet effective framework for MTSF via cross-modality alignment. Specifically, we present a dual-modality encoding with two branches: the time series encoding branch extracts disentangled yet weak time series embeddings, and the LLM-empowered encoding branch wraps the same time series with text as prompts to obtain entangled yet robust prompt embeddings. As a result, such a cross-modality alignment retrieves both disentangled and robust time series embeddings, "the best of two worlds", from the prompt embeddings based on time series and prompt modality similarities. As another key design, to reduce the computational costs from time series with their length textual prompts, we design an effective prompt to encourage the most essential temporal information to be encapsulated in the last token: only the last token is passed to downstream prediction. We further store the last token embeddings to accelerate inference speed. Extensive experiments on eight real datasets demonstrate that TimeCMA outperforms state-of-the-arts.