Attribute Reduction Of Incomplete Information Systems Based On Adjustable Multi-granulation Rough Set

In the process of data processing,data sets are sometimes noisy or default,so theories and methods to deal with inaccurate and uncertain data are necessary.Rough set theory is a new mathematical tool to meet this requirement.At present,rough set theory is widely used in the fields such as particle computing,machine learning,artificial intelligence and data mining.Attribute reduction is one of the most important contents in rough set theory.The existing reduction algorithms of information system mainly include attribute reduction algorithm based on attribute importance,discernibility matrix and information entropy.However,with the progress of computer and data acquisition technology,the accumulation of data is growing rapidly both in number and dimensions.As a result,the complexity of the algorithm is increasing with the increase of data size.Therefore,it is of great significance to develop more effective attribute reduction methods.The global adjustable multi-granulation rough set model is not very available in attribute reduction because of its high computational complexity.In this thesis we present a local adjustable multi-granulation rough set model based on information level,and we define the concepts of lower approximate quality,inner and outer importance of attribute.Finally,we study the difference and relationship between local attribute reduction and global attribute reduction of multi-granulation rough set in incomplete information systems,and a heuristic algorithm is formulated.