Research On Incremental Knowledge Acquisition Algorithm Based On Granular Computing

Knowledge is the carrier of human intelligence. An important researchissue in artificial intelligence (AI) is knowledge acquisition. Research onhuman intelligence is searching knowledge and discovery knowledge. Theuncertainty of AI is a new development trend, the representation andprocessing of uncertain knowledge is the new problem in knowledgeacquisition. At present, the traditional data mining methods can't meetpeople's needs when the data explosive growth. Granular computing datamining method becomes an important data mining method due to its ownadvantages in the process of dealing with complex problems, the purpose ofthe method is to find a better approximate solution instead of the bestprecision solutions, so that the original problem can be simplified and the costwill be reduced.In this paper, based on the granular computing and rough set theory, theknowledge granularity space is established by a hierarchical granularcomputing theory as the basic idea. The details of the paper are as follows,(1) A multi-granularity formal concept analysis method is proposedBased on the granular computing theory. In the method, the whole attributespace is divided into many subspaces, and the purpose is to as much aspossible to reduce the unnecessary data analysis and simplify the originalproblem. First, the whole information system is considered as a coarsegranularity space. And then the multi-granularity attribute description for aformal concept based on the hierarchical granular computing theory andconcept lattice theory is presented. Finally, the correlation of differentgranularity spaces is analyzed. These conclusions are the foundation work forthe following research.(2) Many data mining methods on how to acquire knowledge from thedynamic information system are analyzed in detail at first, and then a newincremental knowledge acquisition method based on granular computing isproposed. In the method, the original information system is considered as acoarse knowledge space, and then the coarse space is divided into manysubspaces, so, a knowledge granule tree is established successfully. When a new data is added into the information system, the corresponding knowledgesubspace will be found at first, and then original knowledge will be quicklyupdated based on rough set theory and Hash data compression method. Theexperiment results show that the method is feasible and effective forprocessing of dynamic data.(3) The incremental knowledge acquisition methods become more andmore important in knowledge acquisition. In these methods, there are twomain ideas. One is increasing new objects, such can be seen as the process ofexpanding the cognition range; the other is increasing new attributes, such canbe seen as the process of deeply understanding of the essential characteristicsof a new thing. A new algorithm is attempted to establish based on both roughset theory and concept lattice theory, in which the nodes are updated orremoved in the process of establishing the hierarchical concept structuremodel. The algorithm is simple and effective.