Research On Theory And Application Of The Full Covering GrC Model

In the age of big data, Internet is producing all kinds of data every momentof our day, how to find the potential and valuable knowledge from thesefast-growing data information rapidly is a hot research spot in data mining.Granular computing (for short, GrC) is a new information processing conceptand computing paradigm, granulation of problems, computing of granules andintegration of the results to granulated problems constitute its basic ideas, whichis a powerful tool to handle imprecise, uncertain and incomplete information,massive data and complex problem. Owing to its unique characteristics,Granular computing method shows great advantages in the data mining.Partial covering (the global GrC model) is an important model under theframework of the GrC theory, which has the potential application values in theresearch fields such as computer security, search engine and customer evaluation.As a special case of the partial covering, the full covering had been studied fromthe perspective of the rough sets, which is depicted under the framework of thegeneral topology. This paper explores the full covering from the perspective ofGrC under the framework of pre-topology theory.1.The concepts of the granule with center, the full covering and the fullcovering approximation space were presented on the basis of neighborhood system, followed by the concepts of interior and closure operators in thepre-topology space. Then the interior and closure operators of the full coveringGrC model were defined by those proposed granules, which could be used toapproximate any object in the full covering approximation space. Thefundamental properties of those two operators were explored under the fullcovering approximation space, the axiomatic systems for the interior and closureoperators of the full covering GrC model were also established.2.The center of the granule was defined which was used to define thegranularity entropy of a full covering and a family of full coverings. Then thesignificance measurement of basic granule and full covering granule werepresented based on the granularity entropy, and the theorems for the reduction ofthe information system which was modeled as a full covering and a family offull coverings were also discussed; then the according knowledge reductionalgorithms based on the full covering GrC model and their complexity weredesigned and analyzed, namely, the reduction of granules in the full covering(RGFC) and the reduction of a covering in a family of full coverings (RCFFC).Finally, examples showed that the proposed algorithms were effective.3.A supplier selection evaluation model was designed based on the fullcovering GrC theory, combined with knowledge reduction algorithm of the fullcovering GrC, a new supplier selection criteria was proposed, which got moreimportant and representative features, and also provide a reference for GrCmodel applying to the data/text mining, customer/credit card assessment and other fields.