EpiRiskNet: incorporating graph structure and static data as prior knowledge for improved time-series forecasting.

EpiRiskNet combines time-series data with graph and static information to enhance forecasting accuracy. This model features the SCI-Block for improved feature extraction and interaction learning, leveraging the capabilities of SCINet and Triformer to manage diverse feature scales. The model's standout attribute, scalability, is driven by Triformer's Patch Attention mechanism, ensuring efficient processing of large-scale data. EpiRiskNet was tested across several locations, including Liaoning, Chongqing, Heilongjiang, and Guangxi, where it demonstrated greater accuracy than other methods. This accuracy is crucial for effectively forecasting disease risks. The model's adaptability to various regional conditions underscores its significance in public health and epidemiology. Moreover, its modular and flexible design makes EpiRiskNet suitable for a wide range of applications that require advanced data processing and predictive analytics. [ABSTRACT FROM AUTHOR]