Rule generation based on rough set theory for text classification

In this paper we describe an approach based on rough set techniques for decision rule generation applied to text classification. A minimal discriminating set - a reduct - for the original data set is achieved through analyzing the degree of dependency among attributes. To speed up the search for reducts, the information gain criterion is used to reduce the number of attributes considered and rank the attributes in decreasing order, and heuristic functions are incorporated into a range of rule generation algorithms.