Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

EpiLLM: Unlocking the Potential of Large Language Models in Epidemic Forecasting

Chenghua Gong, Rui Sun, Yuhao Zheng, Juyuan Zhang, Tianjun Gu, Liming Pan, Linyuan Lv

TL;DR

EpiLLM presents a novel approach to epidemic forecasting by repurposing large language models with a dual-branch token alignment for infection and mobility data, coupled with autoregressive next-token prediction and spatio-temporal prompting. The framework leverages a GNN-based epidemic branch and an MLP-based mobility branch to steer LLMs into a shared token space, enabling accurate, multi-step forecasts while freezing the backbone and training lightweight adapters. Extensive experiments on four real-world COVID-19 datasets show that EpiLLM significantly outperforms traditional baselines and exhibits scaling behavior typical of LLMs. The results highlight the practical potential of LLM-based epidemic forecasters and point to promising directions in prompt learning and efficiency for domain-specific time-series tasks.

Abstract

Advanced epidemic forecasting is critical for enabling precision containment strategies, highlighting its strategic importance for public health security. While recent advances in Large Language Models (LLMs) have demonstrated effectiveness as foundation models for domain-specific tasks, their potential for epidemic forecasting remains largely unexplored. In this paper, we introduce EpiLLM, a novel LLM-based framework tailored for spatio-temporal epidemic forecasting. Considering the key factors in real-world epidemic transmission: infection cases and human mobility, we introduce a dual-branch architecture to achieve fine-grained token-level alignment between such complex epidemic patterns and language tokens for LLM adaptation. To unleash the multi-step forecasting and generalization potential of LLM architectures, we propose an autoregressive modeling paradigm that reformulates the epidemic forecasting task into next-token prediction. To further enhance LLM perception of epidemics, we introduce spatio-temporal prompt learning techniques, which strengthen forecasting capabilities from a data-driven perspective. Extensive experiments show that EpiLLM significantly outperforms existing baselines on real-world COVID-19 datasets and exhibits scaling behavior characteristic of LLMs.

EpiLLM: Unlocking the Potential of Large Language Models in Epidemic Forecasting

TL;DR

EpiLLM presents a novel approach to epidemic forecasting by repurposing large language models with a dual-branch token alignment for infection and mobility data, coupled with autoregressive next-token prediction and spatio-temporal prompting. The framework leverages a GNN-based epidemic branch and an MLP-based mobility branch to steer LLMs into a shared token space, enabling accurate, multi-step forecasts while freezing the backbone and training lightweight adapters. Extensive experiments on four real-world COVID-19 datasets show that EpiLLM significantly outperforms traditional baselines and exhibits scaling behavior typical of LLMs. The results highlight the practical potential of LLM-based epidemic forecasters and point to promising directions in prompt learning and efficiency for domain-specific time-series tasks.

Abstract

Advanced epidemic forecasting is critical for enabling precision containment strategies, highlighting its strategic importance for public health security. While recent advances in Large Language Models (LLMs) have demonstrated effectiveness as foundation models for domain-specific tasks, their potential for epidemic forecasting remains largely unexplored. In this paper, we introduce EpiLLM, a novel LLM-based framework tailored for spatio-temporal epidemic forecasting. Considering the key factors in real-world epidemic transmission: infection cases and human mobility, we introduce a dual-branch architecture to achieve fine-grained token-level alignment between such complex epidemic patterns and language tokens for LLM adaptation. To unleash the multi-step forecasting and generalization potential of LLM architectures, we propose an autoregressive modeling paradigm that reformulates the epidemic forecasting task into next-token prediction. To further enhance LLM perception of epidemics, we introduce spatio-temporal prompt learning techniques, which strengthen forecasting capabilities from a data-driven perspective. Extensive experiments show that EpiLLM significantly outperforms existing baselines on real-world COVID-19 datasets and exhibits scaling behavior characteristic of LLMs.
Paper Structure (46 sections, 14 equations, 7 figures, 7 tables)

This paper contains 46 sections, 14 equations, 7 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Epidemic Modeling
  4. Mechanistic & statistical models
  5. Machine learning models
  6. Deep learning models
  7. LLM for Epidemic Forecasting
  8. Complex reasoning
  9. Time-series prediction
  10. Preliminaries
  11. Problem Definition
  12. Autoregressive Modeling
  13. Methodology
  14. Framework Overview
  15. Dual-branch Token Alignment
  16. ...and 31 more sections

Figures (7)

  • Figure 1: The overall framework of EpiLLM consists of three modules: (1) dual-branch token alignment, (2) autoregressive epidemic modeling, and (3) spatio-temporal prompt learning.
  • Figure 2: Ablation study of EpiLLM for epidemic forecasting.
  • Figure 3: Scaling behavior of EpiLLM on France and Italy datasets.
  • Figure 4: Prompt visualization of EpiLLM.
  • Figure 5: Case study of France (part regions) COVID-19 progression during May 10-12, 2020. Areas shaded in gray denote regions with unavailable surveillance records.
  • ...and 2 more figures