Combinação de previsões: uma abordagem usando wavelets

• Ano de defesa: 2018
• Autor(a) principal: Ahmed, Mohamed Lemine Ould Sid
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Lavras; Programa de Pós-Graduação em Estatística e Experimentação Agropecuária; UFLA; brasil; Departamento de Ciências Exatas
• Palavras-chave em Português: Análise de wavelet; Combinação de previsões; Programação matemática; Séries temporais; Modelagem SARIMA; Vazão de afluentes; SARIMA models; Mathematical programming; Times series; Wavelet decomposition; Flow of atributes; Estatística
• Link de acesso: http://repositorio.ufla.br/jspui/handle/1/29059

The reason behind the importance of accurate forecasting of flow of tributaries lies in the fact that they are crucial to hydrological resources planning, which directly impacts many aspects of the economic and social activities. In this work, a hybrid method of combining times series forecasts is applied. Wavelet analysis is implemented, together with SARIMA models and multi-objective nonlinear mathematical programming. At first, the time series to be analyzed is decomposed via Maximal Overlap Discrete Wavelet Transform (MODWT) up to a certain level of resolution. Subsequently, SARIMA models are used to, individually, model and generate forecasts for every wavelet component. In the last stage, mathematical optimization is used to generate forecasts for the original time series in the form of the optimal linear combination of the wavelet components’ forecasts, previously obtained by SARIMA models. The method was applied to the time series of the monthly flow of tributaries of Samuel’s dam in the state of Rondônia. In terms of predictive gains, pure SARIMA model produced better results.