Metodologia Box & Jenkins: uma aplicação em dados de produção de leite cru do estado de Minas Gerais

• Autor(a) principal: Barbosa, Eduardo Campana
• Data de Publicação: 2020
• Outros Autores: Sáfadi, Thelma, Silva, Carlos Henrique Osório, Manuli, Rômulo César
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Repositório Institucional da UFLA
• Texto Completo: http://repositorio.ufla.br/jspui/handle/1/43223
• topic: Previsão; Modelagem; Tendência; Sazonalidade; Forecasting; Modeling; Trend; Seasonality

The Box & Jenkins methodology was used to obtain a statistical model for estimate the production in liters of milk of the 6 first months of 2013 in Minas Gerais state, adjusting SARIMA (p, d, q) x (P, D, Q)s models, where d and D are the number of differences to remove the trend and seasonality of time series, p and q are the order of the autoregressive and moving average operators, P and Q are the order of the autoregressive and moving average seasonal operators and s is the seasonal periodicity. The Akaike Criterion Information (AIC) procedure was used to select the 6 most parsimonious models and to find the best one the error indicators Mean Squared Error (EQM) and Mean Absolute Percent Error (MAPE) were analyzed, in addition to the assumptions of residues white noise. The Seasonal Autoregressive Integrated Moving Average SARIMA (3,1,2) x (0,1,2)12 was upper, view of the principle of parsimony and with more precise estimates. The forecast was more adjusted to the real values of milk production in Minas Gerais state and the model had smaller error indicators. The residues estimated were by this model white noise.