Study On The Approaches For Updating Approximations Based On Rough Sets Under The Characteristic Relation

The classical rough set theory is based on the complete information system. It requires the completeness and accuracy of the data. Then there are some limitations of its applications in data processing and analysis since data missing often happens in today’s massive data. As an extension of classical rough set theory, the characteristic relation based rough set model takes into account both two cases of missing data,"do not care" and missing values. This model can effectively relieve the limitations of the classical rough set model in practical applications. On the other hand, with the development of information technology, vast amounts of data have been collected and these data are constantly changing. The traditional rough set theory and method can not meet these needs. Therefore, study on the approached of incremental updating algorithm based on extended rough set models has become one of hot research topics.In this paper, we firstly discuss the incrementally methods for updating approximations when the attribute set is unchanged and objects vary under the characteristic relation based rough set model. The variation of characteristic sets is analyzed. Then the method for updating approximation of a given concept (a subset of the universe) is proposed and the corresponding algorithm is developed. Secondly, by the introduction of misclassification probability, the definition of the variable precision characteristic relation based rough set model is given. The relation between the variation of characteristic sets and the misclassification probability is analyzed when the attribute set is unchanged and objects vary. Then the incremental method for updating approximations under this condition is presented. The corresponding algorithm is also designed. Finally, the time complexity analysis of the two incremental algorithms is presented. The runtimes of non-incremental algorithms and incremental algorithms are compared by the simulation. Experimental results veridate that the good performance of the proposed incremental algorithms. It will contribute to improve the efficiency of decision rule acquisition.