Seleção Lazy de Atributos - Uma Nova Perspectiva

Rafael B. PereiraAlexandre PlastinoBianca ZadroznyLuiz Henrique de C. MerschmannAlex A. Freitas

Attribute selection is a data preprocessing step which aims at identifying relevant attributes for the classification task. It attempts to remove from the data set the attributes which do not contribute to, or that can even reduce, the predictive accuracy of a classification algorithm. In this paper, we propose a new attribute selection approach - the 'lazy' approach - which postpones the identification of relevant attributes until an instance is submitted for classifica- tion. Our approach relies on the hypothesis that analyzing the attribute values of the instance to be classified may contribute to identifying the best attributes for that particular instance. Computational experiments are shown and preliminary and promising results are discussed.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: