Metodologia sistemática para avaliar impactos de desastres e de epidemias na população atingida

Abstract

This work aims at developing a new methodology for assessing the impact of an event, such as epidemics and natural disasters, in a given set of places. The proposed method uses machine learning techniques and statistical tools to investigate effects of the event, in one or more variables. Three distinct analyses are performed where two are monovariate and one is multivariate. The Resultant Vectors Graph presents a new technique to visualize results from several control charts in only one diagram and it also allows comparison between data after the event had occurred and its historical limits; the statistical comparison, through paired t tests, allows to compare the change in behavior between possibly affected locations and its control locations; in turn, the multivariate analysis, through Fuzzy c-means clustering algorithm, observes the behavior change in possibly affected locations while investigating the relationship between studied variables. In addition, the methodology presents a way of using the Fuzzy c-means algorithm to determine the set of control locations. Lastly, the proposed methodology is tested in two case studies: the "Fundão dam failure in Mariana - MG" and the "Impact of COVID-19 pandemics in visits for monitoring child growth and development in Brazil".