Geoestatística como método automático de previsão em séries semporais: uma comparação com os métodos de Holt-Winters e Box-Jenkins

Abstract

Time series analysis allows to identify and describe the behavior of a variable over time. This work aims to compare the methods applied to predict the behavior of variables evaluated over time, namely: Geostatistics, Box-Jenkins and Holt-Winters. To achieve the proposed objective, initially, the main theoretical aspects of the methods being compared in this work were studied. Then, the estimation process of the models of the proposed methods applied to the historical series made available by the Blood Banks of the Central Hospitals of Maputo and General José Macamo was developed, regarding the units of blood donated and the HIV prevalence index. Afterwards, a test strategy and selection of more reliable models and consequent methods for a forecast horizon of 12 (twelve) months, of a quantitative nature, were used to support the conclusions presented in this work. The results obtained using this strategy give greater advantage to the automatic method of geostatistics. And in all cases analyzed, the Holt-Winters method proved to be the second ranked in the (rar±ing) of prediction potential. In fact, the ARIMA methods positioned it as Tenvúltimo Tugar except in the series of blood samples infected by the HIV virus collected at the BSHGJM (Blood Bank of the Hospital Geral José Macamo) where it proved to be the best method.