Uma Aplicação De Meta-Aprendizagem Nas Cotações Euro/Dólar

• Ano de defesa: 2018
• Autor(a) principal: Duarte, Felipe De Almeida [UNIFESP]
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de São Paulo (UNIFESP)
• Palavras-chave em Português: Meta-Learnig; Financial Risk; Time Series; Meta-Aprendizagem; Análise De Investimento; Séries Temporais
• Link de acesso: https://sucupira.capes.gov.br/sucupira/public/consultas/coleta/trabalhoConclusao/viewTrabalhoConclusao.jsf?popup=true&id_trabalho=6918970; https://repositorio.unifesp.br/handle/11600/52326

For the investor, knowing when to buy or sell a asset is a decision of extreme importance to make profits in financial markets. However, the same decision-making process, the investor should choose the ideal model for the study and analysis of the data series worked. Due to the large number of models available, choosing an ideal model often makes the task difficult, especially for inexperienced investors. To assist in such a decision, meta-learning can be an ideal tool, by making suggestions of models through applications in similar past data. This tool has already been used for the problem of algorithm selection and presents good results in the selection of study models of time series. In this work, we seek to use the method of analysis and analysis of time series (concrete euro/dollar quotations), through a ranking algorithm that suggests the best models. The results were satisfactory, presenting good predictions for the suggested models.