Uncertainty Measures In Information Systems Knowledge And Its Reduction Methods

In the early 1980's, Z. Pawlak of Poland created a rough set theory, and found a description of the value of the information systems of knowledge of the rough. Since then, information systems and the uncertainty of measurement attribute reduction has always been important in the field of information science research, information on the uncertainty of research and exploration of artificial intelligence run through the course of development for nearly half a century. As the information system can serve as a model of the database, data analysis and decision-making, pattern recognition and machine learning in such areas as many of the problems described in the framework, the creation of rough set theory in the near future on research by experts in a wide range of concerns related research results have been widely and successfully application in data mining, decision analysis, pattern recognition, machine learning and knowledge areas and so on.In this paper, in the framework of rough set theory, based on knowledge is based on the relationship between the equivalent of the division of views, information systems and the measurement uncertainty of attribute reduction issues for further research, particularly on the incomplete information systems with default have been studied and obtained the following results:(1) Considering the impact of similar attributes and missing values in the incomplete information systems, it will be unreasonable if you continue to use the block size to measure the amount of information and roughness of Knowledge. Definite information entropy of fuzzy measures, rough entropy of knowledge and rough sets entropy, prove that it is reasonable and get some character of fuzzy measures rough entropy, then take an example to descript how to choose reasonable measurement to calculation the rough entroy, and apply to the reduction of knowledge in incomplete information systems. It may be provided a new way for measuring knowledge uncertainty and attribute reduction.(2) Correct the error in literature [56-58], improve the attribute reduction based on binary discernibility matrix, and give a counter-example. In this paper, a new measure of attribute significance based on binary discernibility matrix is given, and proves it better than the method in the literature [57].The research of knowlegde uncertainty measurement and attribute reduction is a complex subject, in which there are many issues worthing studying. In this paper some of the research is just a try, there are lots of work to be further studied.