Rough Classification Method Research In Incomplete Decision Information System With Hybrid Value

Along with the unceasing development of information technology and automation , complex and massive data systems are emerged in many fields. These data systems usually hidden lots of useful decision knowledge, which need us to excavate and discover. The rough set theory, proposed by Pawlak in 1982, is a new mathematical tool when processing uncertain knowledge, its remarkable advantage is to analyze data without any prior knowledge. The rough set theory takes the decision information systems as the study object, and obtains the class knowledge or decision rule as an ultimate purpose. In recent years, the research results of theory and application on rough set theory are very abundant. This papar study an even more generalized decision information system, which contain continuous data, discrete data, and missing data. We call this system as an incomplete decision information system with hybrid value.This papar study the incomplete decision information system with hybrid value from three aspects.Firstly, based on the rough set model and method in presented literature, an extension of tolerance relation based on connection degree——a tolerance relation of neighborhood connection degree is presented, then a rough set model based on this tolerance relation——a neighborhood connection degree rough set model is proposed to deal with incomplete decision information system with hybrid value. In addition, we discuss the rough set theory of the maximal tolerance classes based on this tolerance relation of neighborhood connection degree.Secondly, for the given rough set model, combining the attribute reduction theory in presented literature, a new attribute reduction algorithm is proposed in this paper. In addition, we find that the proposed attribute reduction algorithm does not not require to deal with missing values, and can improve the computing efficiency by using presented research results.Finally, In order to reduce the interference of decision caused by redundant attribute, we use the proposed attribute reduction algorithm to incomplete decision information system with hybrid value. After, a couple of classification methods: the classification method based on bayesian decision criterion of minimum error rate combining of neighborhood connection degree rough set model and the classification method based on bayesian cost-sensitive learning combining of neighborhood connection degree rough set model, are given in reduction decision information system. Finally, experiments show that the new methods are objective and feasible.The research results of this paper not only have great value in theory, but also have a certain value in practice.