Predição de aplicação de vacinas baseada em Gradient Boostin Regressor

Abstract

With the advance of science and technology, vaccination has become a crucial tool in public health to prevent infectious diseases. However, optimizing the stock and distribution of vaccine doses is a complex challenge that involves considerations such as shelf life, costs, losses and logistics. Given the relevance of the subject, this study proposes the use of the Gradient Boostin Regressor algorithm to estimate the use of vaccine doses and, consequently, to optimize their distribution and stock, with the aim of minimizing losses due to shelf life and lack of doses for the assisted population. The proposed approach takes into account various relevant factors such as the existence or not of vaccination campaigns, demand, days of the week and the history of distribution of vaccine doses. To validate the proposal, we used the application database of the public health network in the state of Rio Grande do Norte (RN), called RN+VACINA. The approach managed to obtain a Coefficient of Determination (R²) of 0.73 for a five-day advance forecast of the number of vaccine doses to be administered.