The Research For Concept Lattice Construction And Data Mining Application

Data Mining (DM) is a rapidly emerging research field relevant toartificial intelligence and database system in today's digital society, theexplosive growth of many business, government and scientific database havefar outpaced our ability to interpret and digest this data. As a result, thepeople need a new technology that uses it to aquire knowledge andinformation from a lot of datum by automated and intelligent databaseanalysis. DM arises at the historic moment.The DM technology called as one of the three events that affects humanbeing's computer in twenty century, which affects deeply the people's dailylife and production as well as the world's progressing. The immense successof practice is base on the guiding of genuine knowledge. As a result, there areso many experts and learned men are engaged in research of the field indomestic and overseas. At present, the research focus on the fields as follows:GA , association rule mining and so on. But in domestic the research in thedomain was started only before one year or more. For this reason, I am basedon systematic and thorough study of DM technology, initiate a new method todiscovery knowledge by concept ,which inherit and develop DM theory. Thecontent of the whole paper is made up of four parts: introduction;DMsummarize;the research of concept lattice;the research of concept andproposition in database relations. We focus on the following several aspectsin DM: the definition of DM and the current situation;the normal andextended definition of concept;the definition of concept and propositionwith hierarchies relation;the research of data mining of the concept as wellas characteristic ruler with hierarchies relation;initiating the theory of theextent of concept as equivalent. introducing concept lattice;utilizing conceptlattice to express concept and subconcept and equivalent;initiating thetheory Rough concept analysis.(1) Construction of concept lattices. We studied some methods of buildingconcept lattices and sorted out and compared them. The great mass of themneed much time in judging the type of the nodes. A new incrementalalgorithm of building concept lattices was presented in this paper basing onthe relation of nodes. It can save the time of judging the type of the nodes andmend the speed of building concept lattices. At last we compare some of thebest algorithm on concept lattices construction including Gater, Norris,Bordat, Godin and Nourine with the new concept lattices constructionalgorithm on the runtime cost, then it shows that the new constructionalgorithm have a better performance compared with the others.(2) The applications of concept lattices in DM stress studied theconstruction and application of the frequent concept and frequent conceptlattice. Then we studied the relation of concept lattices and association rules.Association rule discovery, as a kernel task of data mining, has been studiedwidely. Concept lattice, induced from a binary relation between object s andfeatures, is a very useful formal tool and has been used in many fields. Itrealizes the unification of concept in tension and concept ex tension,represents the association between objects and features, and reflects therelation ship of generalization and the specialization among concepts, so it isf it for discovering the potential concept below the data. In this paper, therelationship between concept lattice and association rule discovery isanalyzed. Then, the structure of node in lattice is modified according to therequirement, while two algorithms are developed for constructing thecorresponding lattice incrementally and for extracting association rules,where some theorems and properties are used to reduce the number ofdiscovered rules. Finally, the complexity problem is discussed, and thecorresponding experimental result s are given.At last, the application prospect of concept lattices and further researchdomain were presented.