Granular Mining and Rough-Fuzzy Pattern Recognition: A Way to Natural Computation

Rough-fuzzy granular approach in natural computing framework is considered. The concept of rough set theoretic knowledge encoding and the role f-granulation for its improvement are addressed. Some examples of their judicious integration for tasks like case generation, classification/ clustering, feature selection and information measures are described explaining the nature, roles and characteristics of granules used therein. While the method of case generation with variable reduced dimension has merits for mining data sets with large dimension and size, class dependent granulation coupled with neighborhood rough sets for feature selection is efficient in modeling overlapping classes. Image ambiguity measures take into account the fuzziness in grey region, as well as the rough resemblance among nearby grey levels and nearby pixels, and are useful in image analysis. Superiority of rough-fuzzy clustering is illustrated for determining bio-bases in encoding protein sequence for analysis. F-information measures based on fuzzy equivalence partition matrix are effective in selecting relevant genes from micro-array data. Future directions of research, challenges and significance to natural computing are stated. The article includes some of the results published elsewhere. Index Terms — soft computing, granulation, generalized rough sets, rough-fuzzy computing, data mining, bioinformatics, image analysis, case based reasoning.