Modelagem de séries temporais da propagação do Covid-19 utilizando Sarima com parâmetros meteorológicos : um estudo de caso nos 3 maiores centros urbanos de Mato Grosso

Abstract

The corresponding period of the Covid 19 pandemic began at the end of 2019 and lasted for the other years from 2020 to 2022. It arrived in Brazil at the end of February and spread quickly throughout the country, in Mato Grosso, the first case of transmission community took place on 03/19/2020. However, as soon as the competent authorities became aware of the facts, they implemented mitigation strategies and epidemiological surveillance actions, through operational strategies by the health services in the development of their preventive and assistance actions. Thus, the need to understand the evolution of the disease in the state of MT arises, as an instrument to aid actions to combat and control the pandemic. In this way, the present research presents a forecast model employing techniques of modeling of time series data, through the use of epidemiological database of the State Department of Health of Mato Grosso (SES), through the intercession of the use of database data from occurrences of infected people, with meteorological variables (air temperature, relative humidity, and rainfall), as auxiliaries, for the construction of predictive models and to verify how these variables explain and refine the prediction of the trajectory of the virus's advance , with the aim of identifying the structuring mechanism of the series and, therefore, extracting relevant periodicities, detailing its behavior and developing predictive models. The procedures were developed in the R language, based on the evaluation procedures of statistical assumptions, seasonality and validation, a SARIMA model was selected for each city: Cuiabá, ARIMA (4,1,0), the covariance with rainfall obtained the best fit; Várzea Grande, ARIMA (1,1,1), the prediction of data related to the relative humidity variable is the most adequate model; Rondonópolis, ARIMA(1,1,0), the variable "Infected" was the best model to be used to describe the dynamics of the temporal series in the region in the analyzed period. Thus, the application of the model should be interpreted as an aid tool for the actions of epidemiological surveillance services in public policies of government administrations.