Rough‐Fuzzy Granulation and Pattern Classification

Abstract

Granular computing refers to computation and operations performed on information granules, that is, clumps of similar objects or points. This chapter describes a rough-fuzzy granular space (RFGS) using the class dependent (CD) fuzzy granulation and neighborhood-rough-set-based feature selection. It first describes the RFGS-based model for pattern classification. The chapter then presents several quantitative measures such as DS measure, classification accuracy, precision, recall, and kappa coefficient (KC) for completely labeled data sets, and s index and DB index for partially labeled data sets to evaluate the performance of different pattern classification methods. It reports a brief description of different data sets. Finally, the chapter presents different case studies and a comparison among different methods. It uses three classifiers to evaluate the performance of different methods: K-nearest neighbor (K-NN) rule, maximum likelihood (ML) classifier, and multilayer perceptron (MLP). fuzzy set theory; pattern classification; performance evaluation; rough set theory