Research On Shadowed Sets Model Based On Uncertainty Analysis

With the rapid development of information technology,the explosive growth of production data has made the issue of data processing urgent to be solved.How to effectively mine knowledge is worth studying,and how to describe uncertain information effectively is a key issue.Shadowed sets is a three-way approximate model for dealing with uncertainty issues.It can reduce the information redundancy of the fuzzy set by approximately dividing the uncertain objects into three different regions.However,many existing models have large uncertainty loss,that is,fuzzy entropy loss,which means that shadowed set does not keep the consistency of original knowledge information.In addition,many models lack reasonable semantic interpretation for the thresholds.Therefore,based on the in-depth study of the existing models,this thesis proposes two different types of shadowed set extension models.The main research work is as follows:(1)Based on the principle of uncertainty invariance and least cost,mean-entropybased shadowed sets is proposed.Firstly,based on the fuzzy entropy loss,a new decision partition rule is derived and a new threshold is obtained.Then,by analyzing the causes of fuzzy entropy loss,an iterative method is adopted to optimize and expand the proposed model.Finally,through case analysis and experimental analysis,it is verified that the proposed model has a small fuzzy entropy loss after optimization.(2)Based on the compromise principle of game theory,from the perspective of fuzzy entropy,fuzzy-entropy-based game theoretic shadowed sets is proposed.Firstly,the error of fuzzy entropy loss is analyzed,and two kinds of fuzzy entropy loss are analyzed based on the compromise principle.Then,the binary algorithm is introduced to find the threshold pairs quickly.In addition,through the analysis of data sets with different data distribution types,the proposed model is further extended.Finally,through the experimental analysis,it is verified that the proposed model has less fuzzy entropy loss and less time loss.In conclusion,based on different construction principles,this thesis proposes two different shadowed set extension models,the purpose of which is to make the constructed model more effective and have more reasonable semantic interpretation.