Classification Approach Based on Rough Mereology

This article presents a classification approach based on granular computing combined with rough set. The proposed classification approach used the theory of rough mereology and fuzzification in order to classify input datasets into sets of optimized granules. The proposed approach was applied to five datasets of the UC Irvine Machine Learning Repository. The Abalone dataset that consists of 4177 objects and eight attributes was selected as an illustrative example. Empirically obtained experimental results demonstrated that the proposed rough mereology based classification approach obtained better performance compared to other experienced proposed classification approaches.