Model And Algorithm Of Analyzing Data Based On Rough Set Theory

With the rapid development of information technology, the types of data are growing complicated and the scale of it is also on the increase. Resultantly, an enormous amount of high dimensional data come into existence which are complex in type and heterogeneous in form. As it is well known that data are the main carrier of knowledge in fields that are closely linked with our life and production such as social economy, politics, environment and health, it turns out to be a major difficulty in the study of artificial intelligence to know the important information and knowledge which lay undiscovered behind the large amount of complicated data.The data mining technology appears at this imminent requisite time, when mankind dedicates to the acquiring of useful knowledge from data. It is a mul-tidisciplinary field of study, with major tasks including classification, cluster-ing and relevance analysis. Rough set theory is an important method in the field of data mining, characterized mostly for its data analyses like classification an-ticipation and relevance analysis by the information from the data themselves. Being one of the focuses of the study of artificial intelligence theory and its applied fields, this theory presently has already applied successfully in many science and engineering fields.This paper, from the analysis of key problems concerning the rough data analysis, carries out a detailed and systematic research of the theories of neigh-borhood rough set model, fuzzy rough set model and rough set in aspects of uncertainty measure, feature selection algorithm and evaluation. The main re-search findings and innovation are summarized as follows:(1) Based on the in-depth analysis of the neighborhood information gran-ules and fuzzy information granules, the paper has established connection between neighborhood information granular and fuzzy information granular, revealed the relationship among neighborhood rough set model, Hu fuzzy rough set, Wang fuzzy rough set model, Dubois fuzzy rough set model and Radzikowska fuzzy rough set model. In this way, this paper has provided the-oretical foundation for the selection of models in the rough set data analysis, facilitating to the establishment of rough set model。(2) This paper has made the axiomatized definition of roughness mea-sure, defined the changing mechanism of random entropy and fuzzy entropy with partition ordering finding, revealed the inner connection among random, fuzziness and roughness within the framework of rough set theory, provided ways of roughness measure by fuzzy entropy, also put forward the roughness measure based on random entropy and fuzzy entropy, which brings forward binding theory and feasible means for it.(3) Based on the establishment of discernibility matrix of Shannon and Complement entropy, the thesis has made corresponding complete attribute reduction, revealed the inter-relationship of attribute reduction in difference senses and designed to the retaining of roughness. Findings also include the definition of complement in fuzzy approximation space and the fact that changes in fuzzy decision table Shannon entropy are not monotonous. Attribute reduc-tion algorithm based on fuzzy is also found and applied to the hybrid data attribute reduction.(4) Furthermore, the overall certainty measure, consistency measure and support measure with partition changing mechanism have been made clear and the decision performance changes of positive region reduct, Shannon’s entropy reduct and complement entropy analyzed. All these provide grounding for actual application when the selection of proper attribute reduction needs to be made.This paper, with the above-mentioned systematic studies, has attained systemic fruits in the analytical model of rough set, in which the relation-ship between neighborhood and fuzzy rough set has been clarified; the intrin-sic connection of the uncertainty of rough set revealed, the fuzziness measure based on random entropy and fuzzy entropy put forward, a new attribute re-duction algorithm and the means to evaluation brought forward. The findings as such not only enrich and develop the basic theories and means of rough set data processing, but promote technological assistance to the complicated data analysis.