Uncertainty Measurement Of Interval-valued Information Systems And Its Comparative

Uncertainty measurement is a major problem in research of rough set theory.There are two main types of uncertainty measurement:information measurement and algebraic measurement.In this paper,information entropy measurement and accuracy,roughness for interval-valued information systems are constructed.And compared them with the uncertainty measurement are proposed in literature.The specific content are as follows:For uncertainty measurement problem of the lack of a monotonous form of shannon entropy for interval-valued information systems,firstly,a relationship is defined from a given covering in this paper,and easily been proved that it is an equivalence relation.Secondly,a method of deriving a partition from a given covering,and proved that the finer the covering,the finer the partition derived from it,and then the Shannon entropy measure for interval-valued information systems is constructed.Finally,the coentropy(granularity),the joint entropy and conditional entropy are constructed respectively,and some related properties are proved.In view of some uncertainty measurement problems for interval-valued information systems have been proposed in literature,the accuracy and roughness are constructed based on ?-equivalence class in this paper.Compared uncertainty measurement based on ?-equivalence class with uncertainty measurement based on ?-tolerance class in lit-erature,the accuracy based on ?-equivalence class is compared with the accuracy based on ?-tolerance class,the accuracy based on ?-equivalence class is compared with the in-formation entropy,the ?-information entropy proposed is compared with the ?-entropy.Experiments are performed on real data sets,the experimental results show that the uncertainty measurement based on ?-equivalence class is more effective.