Aplicação da metodologia de Box-Jenkins para análise da evolução dos óbitos por neoplasias malignas de pele nas regiões Sul e Sudeste do Brasil

Abstract

In data sets whose observations are ordered in moments of time, possibly at equidistant intervals, the analysis of time series is indicated. This type of analysis allows identifying patterns such as: seasonality, trend, cycles and level changes. In addition, it is possible to generate explanatory models capable of describing the behavior of the sequence of observations, and also to obtain predictions about the analyzed series. In the analysis of univariate time series we have the Box-Jenkins methodology. This allowed us to analyze not only stationary processes using ARMA models, but also non stationary processes through ARIMA models. Therefore, this study proposes to carry out a study using the Box and Jenkins methodology in the series of deaths due to malignant skin neoplasms between 1996 and 2014, in the South and Southeast regions of Brazil, in order to identify patterns, generate explanatory models and obtain forecasts. This study made it possible to show some characteristics such as presence of increasing tendency, absence of seasonality and non-stationary behavior in the series of deaths due to malignant skin neoplasms in both the South and Southeast regions. The Box-Jenkins methodology showed good prediction quality for the series of the South and Southeast regions, which presented ARIMA models (4,1,0) with constant and ARIMA (0,1,1) with constant as best adjusted models to the data. In addition, modeling of the series resulted in adjustment of models with different components, deaths due to malignant skin neoplasms in the South were explained only by AR components and the Southeast region only by MA components, both with integration component.