SIMPLE FEATURE SELECTION NETHODS THAT MAKE A DIFFERENCE: AN APPLICATION TO TEXT CATEGORIZATION

Abstract

Feature selection is an important issue in machine learning. Especially in domains that inherently possess a large number of features, feature selection is a good resort to reduce computational costs. Moreover, gains in accuracy are also expected after feature selection since irrelevant features can act as noise. Text categorization is an area that has been getting a lot of attention lately an applying featuer selection to this domain can be highly beneficial. We presenta very simple approach to feature selection for text categorization and demonstrate that favorable and signigicant results can be obtained using commonsense, rule-of-thumb methods.