EpiSTLLM injects regional adjacency information into Transformer-based representation learning to capture cross-regional transmission structure and long-term temporal dependencies, and achieves the best or highly competitive performance in most evaluation settings against statistical, deep learning, graph-based, and mechanism-guided baselines across 4, 8, and 12-week horizons.
Abstract
Infectious disease outbreaks strain medical resources and public-health systems, making accurate multi-horizon, multi-regional forecasting essential for early warning and resource allocation. Existing approaches face a trade-off: data-driven deep models can capture nonlinear spatiotemporal patterns but often overlook transmission mechanisms, whereas mechanistic models are interpretable but limited in modeling complex regional dependencies. To address this challenge, we propose Epidemic Spatial–Temporal Large Language Model (EpiSTLLM), which is a graph-structure-enhanced large language model for multi-regional infectious disease forecasting. EpiSTLLM injects regional adjacency information into Transformer-based representation learning to capture cross-regional transmission structure and long-term temporal dependencies. A temporal-gated cross-attention module generates horizon-specific latent transmission and recovery parameters, while a latent-space SIR-inspired propagation mechanism with a residual correction branch enables stable multi-horizon forecasting without requiring fully observed compartmental states. Experiments on the FluView state-level influenza-like illness dataset and NHSN state-level influenza hospitalization dataset show that EpiSTLLM achieves the best or highly competitive performance in most evaluation settings against statistical, deep learning, graph-based, and mechanism-guided baselines across 4-, 8-, and 12-week horizons. For example, EpiSTLLM reduces MAE and RMSE values by 9.5% and 6.2% at H=4 on FluView, and by 12.0% and 13.1% at H=8 on NHSN compared with the strongest baselines, respectively.
Reliable epidemic monitoring often requires inferring regional infection burden and transmission heterogeneity from noisy, spatially aggregated, and potentially sparse surveillance data. Agent-based models (ABMs) are attractive for this task because they represent individual behavior, contact heterogeneity, and localized interventions, but these same features make them difficult to calibrate online. We develop a generative AI-based data-assimilation (GenDA) framework for partially observed epidemic ABMs that estimates both the epidemic state and a heterogeneous parameter field while respecting the gap between observable macrostates and latent agent-level microstates. GenDA combines a training-free, score-based generative update for macrostate correction with a direct parameter update based on macrostate discrepancies, followed by a macro-micro reassignment step that restores consistency with the ABM. In controlled and geographically explicit synthetic experiments, the framework recovers regional epidemic burden, dominant hotspot structures, and effective transmission heterogeneity from aggregated observations, while improving post-assimilation forecasts relative to state-only assimilation.
Siming Liang, Jacob Hauck, Minglei Yang et al.· 0 citations
This work proposes MechGNN-Epi, a hybrid framework that couples a spatiotemporal graph encoder with a differentiable SIR update that yields epidemiologically constrained trajectories and produces region- and time-indexed parameter proxies that can be inspected as diagnostic signals, while not being guaranteed as causally identifiable mechanistic parameters.
Modeling spatiotemporal propagation and dynamic regional interactions is essential for intelligent decision-making in large-scale interconnected systems and communication-aware engineering applications. To overcome the limitations of conventional vaccination prediction methods that neglect dynamic spatial dependencies, this study proposes an LSTM-based forecasting framework integrating spatiotemporal graph convolution, dynamic spatial attention, and spatial regularization mechanisms. A dynamic spatial weight matrix combining geographic adjacency and population mobility is first constructed to characterize inter-regional coupling, after which graph convolution and adaptive attention are employed to capture time-varying spillover effects and evolving influence patterns. Experiments conducted on weekly vaccination data from 90 districts and counties in Zhejiang Province demonstrate that the proposed model reduces MAE and RMSE by 19.0% and 17.2%, respectively, compared with the conventional LSTM while significantly improving the spatial consistency of prediction residuals. Furthermore, the framework reveals hierarchical propagation characteristics from core urban clusters to transportation hubs and peripheral regions through interpretable dynamic attention and spillover analysis. By explicitly modeling spatial interaction and propagation mechanisms rather than isolated temporal sequences, the proposed approach provides a generalized methodology for spatiotemporal signal evolution analysis, network-aware prediction, and adaptive information propagation in complex distributed systems.
M. Song, H. Li, L. Yang· Advanced Electromagnetics· 0 citations
Accurate influenza forecasting is essential for public health emergency preparedness and timely resource allocation. Although meteorological factors are established modulators of influenza transmission, existing deep-learning models rarely exploit this physical knowledge in a principled way. We introduce MeteoGST, a meteorology-driven spatiotemporal graph learning framework that combines (i) multi-scale feature extraction across seven operational meteorological variables (T_max, T_min, DTR, absolute humidity q, relative humidity RH, surface-pressure anomaly p_anom, and 10 m wind speed U10) via parallel dilated convolutions, TCN, and Transformer branches; (ii) a meteorology-aware dynamic graph attention network (MeteoGAT) whose edges blend geographic adjacency with time-varying meteorological similarity through a learned gate α; (iii) residual-trend decomposition with a peak-aware composite loss; and (iv) anti-smoothing meta-learning adaptation. On a 34-city pre-COVID-19 benchmark (2018–2019), MeteoGST achieves an RMSE of 0.65/0.85/1.12, MAE of 0.48/0.63/0.82, R2 of 0.84/0.76/0.70, and Peak F1 of 0.72/0.65/0.59 at 7-, 14-, and 30-day horizons, respectively—improvements of 4–6% over the strongest GNN baseline (MPNN-LSTM) and 31–44% over classical baselines (ARIMA/LSTM). Under a 2021–2022 distribution-shift stress test, the model retains its ranking at 1- and 4-week horizons (PCC 0.78/0.62) and degrades gracefully at 8 weeks, demonstrating robustness beyond the training distribution. MeteoGST offers an operationally deployable tool (MeteoGST-Lite: ≈1 ms per city-week on a laptop-class CPU) for integrated meteorology-aware influenza surveillance.
Pengran Qi, Lening Liang, Wenjing Li et al.· Applied Sciences· 0 citations
Influenza-like illness (ILI) remains a persistent global health challenge, necessitating accurate forecasting tools for timely public health response. This study systematically benchmarks fine-tuned large language models (LLMs), e.g., Llama2 and GPT2, for influenza surveillance forecasting in data-limited time-series settings. We develop a lightweight fine-tuning framework that adapts pre-trained LLMs using compact embedding and prediction layers and evaluate it on seven weekly aggregated real-world surveillance datasets. Despite sample sizes of only ∼523 time points per region and the absence of cloud-based data transfer, fine-tuned LLMs consistently outperform SARIMA, LSTM, PatchTST, CoVTransformer, FEDformer, Time-LLM, and GPT4TS in both accuracy and stability, especially for long-term forecasts across diverse geographic settings. Even in zero-shot settings, pre-trained LLMs capture broad epidemic trends with performance comparable to SARIMA. These findings establish fine-tuned LLMs as efficient and robust forecasting tools suitable for privacy-sensitive, data-scarce public health applications.
A unified complete-data likelihood framework for epidemic processes evolving on partially observed dynamic networks that contributes to statistical inference for partially observed interacting stochastic systems on evolving networks and establishes a foundation for uncertainty-aware analysis of complex transmission processes.