Refining decision tree classifiers using rough set tools

Abstract

The proposed hybridized rough set framework is composed of traditional Rough Set (RS) approach and classical Decision Tree (DT) induction algorithm. RS helps to identify dominant attributes and DT algorithm results in simpler and generalized classifier. The implementation of the Hybridized Rough Set Framework is presented as the RDT algorithm. GA heuristics are used to generalize the RDT algorithm further. Experimental results obtained on applying the hybridized rough set framework and the related base algorithms on datasets belonging to the three categories are presented in this paper. Accuracy, complexity, number of rules and number of attributes in the induced classifier assess the performance of the candidate algorithms. The results indicate that the proposed framework is an effective model for classification.