Research Of Data Reduction And Application Based On Rough Sets

Data mining is to extract and analyze some subtle relationships inside the enormous data sets, and then comes to a useful and understandable conclusion. The theory of rough sets is a mathematical tool for processing ambiguous and inaccurate problem. It is well-known that multitudes of practical problems can't be solved by the present computer systems, especially those like machine learning, pattern recognition and artificial intelligence, which make the algorithm for describing individuals impossible. The theory of rough sets and its expansion can provide an accurate mathematical technology for rough description of such individuals.With development of knowledge economy and the coming of information society, more and more information accumulates within its system. Solving the problem of information expansion is not only the research subject of the system itself, but also an important study field on the Internet. The reduction of information system can reduce the information volume by removing irrelevant and unnecessary information without affecting its functions. After reduction, the information will be rearranged and new rules obtained. These rules different from the previous ones in terms of their premise and conclusion information. However, the same or similar results can be achieved through deduction.This thesis mainly introduces the basic theories and present researches of rough sets, analyses data reduction, the core of data-mining process, and suggests a data reduction algorithm of informational system, based on Binary Trees. This algorithm makes attribute reduction consistent with its value reduction, shortens the reduction time , and lowers the complication degrees of time and space.The major achievements and innovations made in this thesis are as follows:(1) The present data reduction algorithm is studied, which needs to go through all data in the decision-form to lower the complication degrees of time and space.(2) Based on Binary Trees, a new data reduction algorithm is put forward, including attribute reduction and its value reduction. With the help of the new Binary Trees invented from the dynamic synthesis of revolving function.(3) A prototype system of data reduction algorithm is devised, based on the structure of Binary Trees. The validness and high efficiency of this algorithm is proved through testing several standard test data sets provided by UCI, and it has been uesed in the teaching evaluation system.