Study Of Knowledge Acquisition Based On Binary Discernibility Matrix

With the rapid development of science and technology, the growing amount of data stored in thedatabase, and a lot of data is often incomplete indistinguishable, obtaining implicit rule from a lot ofmessy data is becoming increasingly important. This makes knowledge acquisition has become ahot research issue. As knowledge reduction technology becomes more sophisticated, manyknowledge acquisition algorithms combine with rough set theory, binary discernibility matrix forthe indistinguishable, incompatible, fuzzy uncertain information existing in data. Research based onrough set theory and binary discernibility matrix to deal with incomplete data sets which areincompatible, includes the following aspects:1、Propose an improved attribute computing core algorithm based on binary discernibilitymatrix. After the analysis of attribute computing core algorithm, found the problem that computingcore based on binary discernibility matrix is inaccurate. That is, using the total number of1in a rowof binary discernibility matrix as the sole criterion to measure and determine the core of theattributes, will inevitably causes the errors of core result in dealing with incomplete informationsystems. The improved computing core method uses a uniform threshold to filter out the attribute ofa relatively small probability to become a core attribute. Experimental results show that thealgorithm has better reduction performance, improves the accuracy of attribute reduction.2、Propose a generalized binary discernibility matrix concept and definition. The previousknowledge reduction focuses on complete information systems which based on the classical roughset theory. For incomplete information systems, a variety of extended rough set models areproposed, however, using the extended rough set models to attribute relative reduction is still nouniform definition and relatively mature knowledge reduction methods. Generalized binarydiscernibility matrix is the revise of the traditional binary discernibility matrix, applicable for theindiscernible relation of complete decision table and extended rough set models of incompletedecision table, able to make the algorithm more flexibility and provides the implementation methodfor the subsequent knowledge reduction and rule extraction.3、Propose an algorithm of knowledge reduction and rules extraction based on the generalizedbinary discernibility matrix. Based on the analysis of rough set models and binary discernibilitymatrix, found that complete decision table is ideal handling objects for classical rough set theory,however, fuzzy uncertainty incompatible phenomenon of data objects is very common. Generalized binary discernibility matrix can be used for attribute reduction and rule extraction for complete andincomplete decision table, consider the case of compatibility and incompatibility. It’s more accurate,more flexible. Through the simulation of UCI data, it has verified the validity of the generalizedbinary discernibility matrix used for attribute reduction algorithm of complete and incompleteinformation system.In summary, the research of knowledge reduction algorithm based on rough sets and binarydiscernibility matrix, provides favorable support for dealing with the problem of incompatibilityincompleteness in knowledge reduction algorithm. It has a good theoretical value and significance.