Avaliação da precisão dos métodos ARIMA, Suavização Exponencial e Redes Neurais na previsão de Séries Temporais Anuais brasileiras

• Autor(a) principal: Camilo, Gustavo Inocencio
• Data de Publicação: 2019
• Tipo de documento: Trabalho de conclusão de curso
• Idioma: por
• Título da fonte: Repositório Institucional da UFU
• Texto Completo: https://repositorio.ufu.br/handle/123456789/25914
• topic: Séries Temporais Univariadas; Métodos de Previsão; Comparação de Métodos; Erros de Previsão.; Univariate Time Series; Forecast Methods; Comparison of Methods; Forecast Errors; CNPQ::CIENCIAS SOCIAIS APLICADAS

Among the applications of Time Series Analysis, forecasts of future values is one of the most used. In forecasting, the precision of the method used is one of the critical factors in its adoption. Three of the major forecasting methods available, due to the widespread use and quality of predictions, are exponential smoothing, ARIMA model, and artificial neural networks. This article aims to test the hypothesis of equality or not between the precisions of point forecasts generated by these three methods when applied to contemporaneous Brazilian time series. For this purpose, 40 annual series of IPEA were selected by means of simple random sampling, predictions with the three methods were executed and their precisions were calculated. The conclusion was that the three methods have the same degree of accuracy on average.