Análise multiescala da propagação de epidemias utilizando regressão por processo gaussiano.

Abstract

The exponential growth in the number of COVID-19 infections has impacted millions of lives around the world, as reported by various news outlets and data dissemination plat forms. Analyzing such data can support the prediction of disease behavior at different scales, aiding decision-making on containment measures across various territorial levels. In this context, this study investigates the application of different modeling approaches used in epidemiological research, aiming to identify the most suitable one for understanding the dynamics of epidemic spread in a multiscale analysis. To this end, the performance of com partmental models, additive regression models (using the Prophet software), and Gaussian Process Regression was evaluated, with Gaussian Process Regression being adopted as the main focus of the research. As its main contribution, the study proposes a deep learning based approach to optimize the kernel selection process in Gaussian Process Regression models, addressing a well-known limitation in the literature and enhancing both predictive performance and computational efficiency.