Local Rough Set Based On Complete Information System

This article mainly discusses based on the local rough set model and the local multi-granular rough set model to further expand the local rough set theory.Based on the local rough set model,considering that decision data is often uncoordinated,this paper introduces it into uncoordinated information systems.The study found that,based on the local rough set of the inconsistent information system,there is a certain relationship between the correct classification rate and the classification quality: different correct classification rates correspond to different classification quality,and the classification quality between the two correct classification rates is maintained Invariable;attribute reduction defined by whether the lower approximate distribution changes or not,there is an attribute kernel,and the reduction is stable,based on this,this paper defines the lower approximate distribution reduction.According to the decision table,the information entropy of the conditional attribute set has the property of monotonous decline.In this paper,the information entropy is used to measure the importance of the attribute,and an attribute reduction algorithm based on the local rough set of the uncoordinated information system is designed.Based on the local multi-granularity rough set model,this paper gives the local multi-granularity rough set model under the two fusion strategies of "seeking common ground while reserving differences" and "seeking common ground while repelling",respectively defining the upper and lower approximate operators under the two fusion strategies,and The corresponding properties are studied.This paper finds that,according to the definition of approximate quality given in this paper,there is a relationship between the correct classification rate and the classification quality similar to the local rough set under the uncoordinated information system: different correct classification rates correspond to different classification quality,two The classification quality between the correct classification rates remains unchanged.In the local multi-granularity framework under the two fusion strategies,with the monotonous change of the granularity space,the approximate quality of the local multi-granular rough set also changes monotonously,and can fully represent the change of the lower approximate distribution in the local multi-granular decision space.Using this feature,this paper defines the particle size entropy based on the approximate quality,gives a reasonable measure of the internal andexternal importance of the particle size according to the particle size entropy,and designs a particle size reduction algorithm based on the lower approximate distribution.The research shows that under the pessimistic fusion strategy,with the increase of the granularity in the granularity set,the lower approximate distribution of the granularity set gradually decreases.In the optimistic fusion strategy,as the granularity of the granularity increases,the lower approximated distribution of the granularity The entropy gradually increases.When the internal importance of a certain granularity is greater than zero,the granularity at this time belongs to the granularity kernel.Finally,this paper verifies the effectiveness of the algorithm through experiments.