Rough Sets In The Mass Of Scientific Data Mining Applications

With accelerating pace of science development, the demands on capability of large-scale information analysis are increasing. Data Mining is the process through which we can extract hidden but useful knowledge from half-backed, noising, fuzzy and random data sets. In order to obtain satisfying results, it is necessary for the database to be preprocessed before the actual mining. Preprocessing converts the original dirty data into normalized data. Data normalization and attribute reduction are the core of the whole process. How to remove redundant attributes is highlighted in both technology and theory research. In recent few years, Rough Set theory has shown potential in attribute reduction. Comprehensive researches and tests reveal that applying Rough Set method on scientific data to reduce dimension can improve the efficiency of pattern recognition. Scientific data has the characteristic of high dimensional, incomplete, noising, and so on. Our purpose is to develop an open, easily maintainable, extensible and user-friendly Scientific Data Mining System(SDMS) which can extract useful knowledge from scientific data and analyze the simulation result of scientific computing. The system is hosted on DBMS and Windows platform. Currently a prototype system with functions of attribute reduction, discretion, classifying and clustering has been completed. Test results show that the system has basically achieved the designing requirement. This paper mainly discusses topics on solving attribute reduction problems by applying rough set methods in the field of scientific data mining. First, basic concepts of Data Ming are introduced; then the theory of Rough Set and its application on attribute reduction is provided; in the 4th chapter, the concept of entropy and the application of entropy combining with Rough Set is recounted; some common attribute reducing algorithms...