Research On Rough Sets Theory Based Data Mining

With the rapid development of information technology such as computer, network, communication and so on, the increase of information takes on going up beyond the exponential speed. The mechanism of searches and query of traditional databases and the method of statistical analysis greatly cannot meet the realistic demand with the information sharp increasing. Lots of data is outdated before its analysis. And it is too difficult to analyze the relations among a great deal of data because the data is too much. It has become research hotspot in many fields that how not only ostensible but also embedded knowledge and information are mined from a great deal of data. In the background, the new technology of data processing, that is Knowledge Discovery in Database, is produced.Knowledge discovery in databases is the nontrivial process of identifying valid, novel, potentially useful and ultimately understandable patterns in databases. Data Mining is the core step during the course of Knowledge Discovery in database. At present, it is a quite active research field.The theory of Rough Sets, presented in 1982 by Polish mathematician Pawlak Z, is a powerful mathematical tool for analyzing uncertain, fuzzy knowledge. Rough sets, as a new hotspot in the field of artificial intelligence, can effectively deal with the expression and deduction of incomplete, uncertain knowledge. The theory of Rough Sets is specially fit for the application to Data-Mining because of its features. Now the method of Data-Mining based on Rough Sets has become one of the main methods of Data-Mining. The study on Rough Sets based Data Mining has greatly theoretical and realistic meaning.The correlative theory of Rough Sets and Data Mining was delivered in this dissertation. We presented a kind of expanding model of Rough Sets, that is the model of Rough Sets with the grade of membership and weight, after lucubrating the deficiencies of the theory of traditional Rough Sets. In this model, we dissertated the information system with the grade of membership and weight, and researched into the process of noise, the partition of approximate space, the calculation of the dependent grade of decision-making attribute to conditional ones, the attributes reduction, the construction of excavating step of correlative rules etc. And the modelis feasible through the validation of an example. This expanding model of Rough Sets overcomes the deficiencies that its classification is too strict and it is excessively sensitive to the noise and some rules kept in boundary are lost etc. as far as traditional Rough Sets is concerned. This model completely succeeds the characters of Rough Sets and holds its all strongpoints. It provides a method that is commonly used in statistic and applied to more objects being classified on the condition of a given error ratio. It will obtain better application in some aspects such as analysis of information system, artificial intelligence and its application, decision support system, knowledge discovery in database, pattern recognition, classification and fault diagnosis etc.For the future, realistic soft system based on this model of Rough Sets will be theoretically lucubrated and exploited.