Forecasting dengue cases through time-series modeling with Google Trends and deep neural networks

Abstract

Dengue fever remains a persistent public health challenge in tropical regions, characterized by complex transmission dynamics and nonlinear outbreak patterns. In this study, we propose a data-driven forecasting framework that fuses real-time public interest signals captured through Google Trends, with a suite of advanced deep learning models. Our approach leverages the inherent nonlinearities in online search behavior to anticipate weekly dengue incidence, achieving state-of-the-art performance across multiple forecasting horizons. Remarkably, we find that a single search term (“dengue”) exhibits strong predictive power, outperforming multivariate feature sets in several models. The findings highlight the potential of low-cost, population-level digital traces as proxies for epidemiological signals and offer a practical, interpretable, and scalable methodology for early outbreak detection in complex systems.