Comparação do desempenho dos modelos estatísticos aplicados a séries temporais de umidade relativa do ar

Abstract

RESUMOModelos estatísticos aplicados a séries temporais vêm sendo empregados em diversas áreas e têm surgido como necessidade atual para empresas sobreviverem em um mercado globalizado e competitivo. ABSTRACTStatistical models applied to time series have been used in several areas and have emerged as a current need for companies to survive in a globalized and competitive market. In this context, the present work aims to present a comparison of the performance of the statistical models: Seasonal Moving Average Self-Correlation (SARMA), Self-Regressive Moving Average and Entry of Exogenous Variables (ARMAX), and additive / multiplicative Holt-Winters exponential smoothing models, adjusted to relative air humidity (RH) data.. Data were collected from the National Institute of Meteorology (INMET) from October 1, 2001 to January 22, 2014. In the comparison of the results and in the selection of the best model, the criteria Absolute Average Percent Error (MAPE) were used, Mean Square Error (EQM), Absolute Mean of Errors (MAD) and Sum of Squares of Differences (SSE). In the analyzed period, it was observed that the mean RH was 76.03% with a coefficient of variation of 7.68% and the series under study presented seasonality, justifying the use of the proposed models. The results showed that the ARMAX (3.0) model with the inclusion of exogenous variables was adjusted