Research On Evidence Synthesis Consistency And Three-way Decision Relationship

With the development of information epoch,people are receiving and processing huge amounts of data from a wide range of sources and in different forms every day.It is a big challenge that how to purposefully extract the information needed from the data and fuse it into realistic,guiding decisions.Evidence theory is an important tool in the field of information fusion because of its advantages in dealing with uncertainty problems caused by ambiguity and randomness.In this thesis,the consistency of evidence synthesis and three-way decision relationships of evidence are investigated.The first study is carried out on the consistency of evidence synthesis.In the division-based trust structure,the evidence synthesis formulas in different cases are discussed based on the information granule perspective,and the conditions for satisfying the consistency of the evidence synthesis are illustrated and proved.In the coverage-based trust structure,it is shown that the intersection operation of coverage is not consistent with the focal elements obtained by the evidence combination operation in the subdivision case.And the conditions for satisfying the consistency of evidence synthesis in different cases are explored and proved.Secondly,the three-way decision relationships of evidence are investigated.By combining the idea of three-way decision,the three-way decision relationship model is established.This model calculates the consistent probability by focal element similarity,sets the cost function by using Euclidean distance,and determines the evidence relationship based on the principle of minimum risk.The results of the examples show that the model can accurately and appropriately determine the evidence relationships.Finally,based on the three-way evidence relationship model,a new multi-source evidence fusion strategy is proposed.This strategy merges evidence with consistency through evidence sets,distributes the evidence to the inner or boundary of different evidence sets by using a three-way decision relationship model,and finally corrects and fuses the evidence according to its distribution.The calculating samples shows that the method not only can be of robustness when dealing with evidence paradoxes compared with other methods,but also has some advantages in reducing the computation and improving the fusion effect.In this thesis,we study the consistency of evidence synthesis and the three-way decision relationships of evidence,and apply the three-way decision relationship model into the process of evidence fusion,which promotes the fusion of evidence theory and other theories,and also further enriches and expands the evidence theory system.