Integrating environmental clustering to enhance epidemic forecasting with machine learning models

Abstract

The COVID-19 pandemic underscored the urgent need for more accurate and adaptive forecasting models to support public health decision-making and limit disease spread. However, many existing models overlook the influence of environmental and climatic factors that significantly affect transmission dynamics. This study addresses this gap with a novel forecasting framework that integrates environmental data into predictive modelling. Our key contributions are threefold: (1) we analyse the relationship between environmental variables (temperature, humidity, and air quality) and COVID-19 trends across countries; (2) we propose a two-stage approach combining K-means clustering to group countries based on environmental conditions, followed by region-specific machine learning models using Support Vector Regression (SVR), Prophet, and Long Short-Term Memory (LSTM) networks for both univariate and multivariate time series forecasting; and (3) we demonstrate that LSTM significantly outperforms other models, achieving superior accuracy for 30-day COVID-19 case predictions. Our results highlight the importance of incorporating environmental variables in epidemic modelling and offer a practical tool for more targeted and effective public health responses. This research provides actionable insights that can inform the design of climate-aware forecasting systems for future pandemic preparedness.