A Comprehensive Survey on various Feature Selection Methods to Categorize Text Documents
Е. С. Брискин, Mohnish Karthikeyan B · International Journal of Computer Applications · 2017
Feature selection is one of the well known solution to high dimensionality problem of text categorization.In text categorization, selection of good features (terms) plays a very important role.Feature selection is a strategy that can be used to improve categorization accuracy, effectiveness and computational efficiency.This paper presents an empirical study of most widely used feature selection methods viz.Term Frequency-Inverse Document Frequency (tf•idf ), Information Gain (IG), Mutual Information(MI), CHI-Square (χ 2 ), Ambiguity Measure (AM), Term Strength (TS), Term Frequency-Relevance Frequency (tf•rf ) and Symbolic Feature Selection (SFS) with five different classifiers (Nave Bayes, K-Nearest Neighbor, Centroid Based Classifier, Support Vector Machine and Symbolic Classifier).Experimentations are carried out on standard bench mark datasets like Reuters-21578, 20-Newsgroups and 4 University dataset.