Análisis de series temporales de casos confirmados y fallecidos por Covid – 19 del Ecuador

Abstract

The objective of this work was to model curves of confirmed cases and deceased people of the COVID-19 disease, represented by the accumulated number of cases as a function of time, Ecuador 2020. The research was quantitative in nature, object applied with a level of explanatory deepening, inductive inferential type with longitudinal cut. The analysis was performed using data from 168,192 confirmed cases and 8,357 verified deaths from the 24 provinces of Ecuador, from March 13 to October 31, 2020, information that is reported by the Ministry of Public Health of Ecuador on its open data web page. The main results of the exploratory analysis of the time series was the identification of increasing trend, absence of seasonality and stationarity, serial models were obtained based on information criteria and measures of adequacy for the series of confirmed cases ARIMA (0.2 , 2) and death cases ARIMA (1,1,1) with drift, the residuals of the two models comply with the assumption of stationarity and white noise but do not fulfill the assumption of normality, the Bootstrap methodology was based on the statistical method of ARIMA prediction, the performance of the methodologies was evaluated by comparing the actual values and the forecast for the months of November and December. Finally, it was determined that the ARIMA modeling is adequate to forecast confirmed cases and the Bootstrap modeling for deceased cases, it is verified that it is not possible to choose only one method of modeling due to the performance of the models and considering that they are time series with different characteristics. It is recommended to contrast the existence of more accurate forecasts of confirmed and deceased cases by means of different statistical techniques such as: Holt Winter, decision trees, artificial neural networks, among others, suitable for working with the accumulated number of cases.