Research On Rough Sets Theory And Its Applications In Knowledge Acquisition

Rough Sets (RS) theory, introduced by Pawlak Z., is a new mathematical tool to deal with knowledge, particularly when knowledge is imprecise or inconsistent. The RS theory gives a formal definition of knowledge so that the knowledge can be analyzed and manipulated effectively. This theory also provides a suit of tools, i.e. reduction of knowledge, to acquire knowledge from data automatically. Recently, the RS theory is widely being used in many areas, such as artificial intelligence, pattern recognition etc.Based on the characters of RS theory, it can be used in knowledge acquisition to support such steps as data pretreatment, data reduction, rule generation and acquisition of data dependencies. The details were studied as follows:In chapter one: The summarization of RS theory was introduced firstly, which included four parts: basic concepts of RS theory, characters of RS theory, application software commonly used in RS theory and research fields of RS theory including theoretical and applied fields. Then, knowledge acquisition based on RS theory was briefly introduced. At last, the content and structure of the dissertation were brought forth.In chapter two: The relationship between information and roughness of knowledge in RS theory was discussed in detail based on the concepts of information entropy and mutual information of knowledge. The roughness of knowledge was quantificationally described with information. Then, the new expression of main concepts in RS theory was given based on information, which was called information expression. Also, the intuitionistic signification and rationality of this expression were explained.In chapter three: A method for evaluating the significance of information system attributes based on RS theory was proposed. A parameter a R(X) defined by the concepts of lower and upper approximation in RS theory was used to estimate the significance of system attributes. More information about the studied system can be found by this method compared with others.In chapter four: The meaning, steps, classification and several existing methods of discretization of attribute in information system were introduced firstly. Then, An area-independent automatic approach for discretization of continuous attribute based on dynamic cluster algorithm is presented. The comparisons between this approach and other exiting methods were made, and the results were satisfying.In chapter five: The reduction algorithms of information system were studied. By studying discretization methods and attributes reduction based on RS theory, a method for rules reduction to real value information system was proposed. By researching fuzzy sets with probability method, a approach for rules reduction to fuzzy value information system was brought forth based on RS theory. Combining concept of support in association rules and RS theory, RSVR algorithm was presented. Finally, the above approaches were tested with several databases.In chapter six: The actualities, research fields and problems of information fusion were generally stated and the approach to multisensor fusion with RS theory was discussed. It can improve the fusion speed and decision ability of the system.In chapter seven: All of the work in this dissertation was summed up, and the future researches in this area were prospected.