An Extensive Empirical Study of Feature Selection for Text Categorization

We present a novel feature selection (FS) approach for text categorization. It first constructs a local feature set for each category by selecting a set of features based on three different schemes: DF, TF and TFIDF, and then constructs a global feature set utilizing well-known CHI method based on the local feature set. The experimental comparison is carried out between our method and CHI method. Results from the experiments are summarized. The results show that our proposed method based on DF scheme can perform comparatively well with CHI methods.