Estudo da regressão por processo Gaussiano para previsão da evolução dos casos de COVID-19 em Campina Grande - PB.

Abstract

Current expectations involving Artificial Intelligence and Machine Learning are high — both in the scientific community and in the industry. The idea that we can extract complex functions from a large amount of data is inspiring and leads us to believe that we can improve our ability to predict future events and make better decisions using a data-driven methodology. Within this context, we can use the Gaussian Regression Process algorithm to perform supervised learning tasks in time series. It is a non-linear and non-parametric regression algorithm based on Bayesian learning. In this we study the main characteristics of the Gaussian Regression Process, focusing on the properties of its kernel functions. 3 basic problems are analyzed allowing us to visualize the algorithm training steps and the distinctions between such kernel functions. Then, we apply it in practical problem to predict the evolution of Covid-19 cases in the city of Campina Grande - PB.