Research On Incremental Knowledge Reduction Algorithm For Imcomplete Information System

Rough set theory has obvious advantages in dealing with incomplete,inaccurate,inconsistent data and other uncertainty data,which have been widely used as data analysis tool.Nowadays,with the emergence of the massive dynamic data,the features that data is incomplete and dynamic become more and more obviously.Incremental reduction in incomplete information system is of great significance for incomplete dynamic data processing.According to whether the decision tables have been completed first,this paper took different methods to deal with the dynamic decision tables attribute reduction.The main work includes three aspects:(1)Completion of incomplete information systemAt present,data completion methods based on rough sets mostly compute the similarities between the object that contains missing values and other object that does not contain missing values,and then use the values of the most similar object to complete the missing values.However,the problem in these methods is that all the condition attributes are considered as equally important,and they ignore the differences between condition attributes.Given this problem,a new notion of fuzzy weighted similarity is introduced,and the similarities between different objects are computed based on the dependencies of decision attribute on condition attributes and the significances of condition attributes.Moreover,the data completion method with rought sets based on the measurement of fuzzy weighted similarity is proposed.With the computing of example and being compared with the current data completion algorithms,the validity of the method is proved.(2)Incremental attribute reduction of incomplete information system based on data completion.The attribute reduction result should be updated continually with the dynamic changing of data in decision table,but how to get more optimized result in the less computing time is still a hot issue.For the completed decision table,firstly,an improved algorithm is introduced to acquire a simplified decision table which is equivalent to the original decision table.And then,combining the merits of the discernibility matrix and positive region,an incremental updating algorithm for attribute reduction,in which just the discernibility matrix element set produced by the inserted new objects is needed to store,is further developed under considering the reduction result of original decision table.Theoretical analysis and experimentalresults show that the proposed algorithm can quickly update attribute reduction result on the basis of the original attribute reduction result when decision tables change dynamically(3)Incremental attribute reduction of incomplete information systems based on the extended model of rough setThe classical rough set theory is constructed on the basis of strict equivalence relation which is based on the assumption that information systems are complete.the integrity and accuracy of data was highly required by this strict equivalence relatio n method.However,nowadays,the incomplete and dynamic increment data often occur indata processing.After analyzing the existing extended model of rough set,an incremental attribute reduction algorithm for imcomplete information system based on positive region is proposed.Theoretical analysis and experimental results show that the proposed algorithm can quickly update attribute reduction result on the basis of the original attribute reduction result.