Abstract

The COVID-19 pandemic highlighted the importance of spatio-temporal epidemiology prediction, which utilizes both spatial and temporal data to forecast disease dynamics. This approach is critical for public health management, particularly during pandemics, due to its ability to analyze disease spread patterns over time and across different regions. Traditional data sources provide extensive spatio-temporal datasets, yet they underscore a need for sophisticated predictive tools that are accessible to public health researchers without deep technical expertise. To address this, we introduce DeepEST, a Python library designed to facilitate deep learning-based spatio-temporal epidemiology prediction. DeepEST integrates advanced predictive modeling techniques such as graph neural networks and recurrent neural networks with traditional epidemiological models, offering a comprehensive suite of tools for data preprocessing, model development, and visualization. This library simplifies the development and application of predictive models, reducing the technical barriers for researchers and enhancing the capability of the public health community to effectively respond to epidemiological threats. This paper details the development and capabilities of DeepEST, showcasing its potential to democratize epidemiological research and public health response strategies. Our code is available at https://github.com/v1xerunt/DeepEST.