A Feature Selection Method Based On The Principle Findings Of Granularity Partitioning And Granularity Overlapping

Granular Computing is the new concept and computing model to informationproceeding, and then the main idea is problem solving on different granualr hierarchies;Granular Computing is theory which can effectively analyse and solve fuzzy, imprecise,inconsistent, partial true; Granular Computing covers all theories, methodologies,techniques, and tools which make use of granules. Granular Computing is not only thesuperset of fuzzy set theroy, rough set theroy, quotient space theroy, but also have becomean important research branch of artifical intelligence field. And granular computing has abroad application prospects.First, in the theory of granular computing and formal concept analysis theory, thepaper discusses the basic concepts related to granular, granularity, granulation, partialorder and formal context. In the formal context of Traditional Chinese Medicine syndrome,syndrome elements, people can construct granular and granular space in order to simplifythe concept lattice. According to the structure of granular and granular space, the paperdisscusses an attribute reduction method based on Granular Computing which not onlycan meet people’s needs, but aslo hold complicated background, clear level, simplification.By this method granular has become a hierarchical relationship. For example, forattributes projective space, the finest granular space is decided by attribute M , and everyconcept is a granular; The thickest granular space is made up of attribute , and all of theconcepts is a granular.Second, the paper presents a new method which is related to the optimization of theformal context and the construction of the attribute partial order graph. In order to conducta granularity devision graph which is no cross, clear level and have thick or finegranularity.Finally, the paper is based on the actual examples of nephrotic syndrome to constructattribute partial order graph. In other words, attribute partial order graph reflects thehierarchical of the granularity and the construction of equivalence relations, forconducting the diagnose of the nephrotic syndrome or rule extraction.