Aplicação de modelos gráficos probabilísticos computacionais em economia

• Ano de defesa: 2009
• Autor(a) principal: Colla, Ernesto Coutinho
• Orientador(a): Orellano, Verônica Inês Fernandez
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Palavras-chave em Português: Modelos gráficos probabilísticos; Redes bayesianas; Risco-país; Modelos macroeconômicos
• Palavras-chave em Inglês: Pattern recognition; Machine learning; Data mining
• Link de acesso: http://hdl.handle.net/10438/4261

We develop a probabilistic model using Machine Learning tools to classify the trend of the Brazilian country risk expressed EMBI+ (Emerging Markets Bond Index Plus). The main goal is verify if Machine Learning is useful to build economic models which could be used as reasoning tools under uncertainty. Specifically we use Bayesian Networks to perform pattern recognition in observed macroeconomics and financial data. The results are promising. We get the main expected theoretical relationship between country risk and economic variables, as well as international economic context and market expectations.