Previsão de retornos intradiários através de regressões usando funções-núcleo
Carregando...
Arquivos
Data
2009-06-10
Autores
Orientador(res)
Métricas
Título da Revista
ISSN da Revista
Título de Volume
Resumo
The contributions of this paper are twofold. First we discuss and apply a method for the evaluation of non linear regressions in forecasting intraday returns of Brazilian stocks, in order to maximize the return of a simulated trading portfolio. Second, Kernel regressions associated with Nearest Neighbors sample partitioning are carried out. Some independent variables are technical indicators, which parameters are optimized in-the-sample. The results are positive as a trading strategy and outperformed by a small difference the linear autoregression benchmark model in a quartile per quartile analysis