Research On The Improvement Of Evidence Theory Fusion Method

Uncertainty is often accompanied in human cognitive activities,but the expression of uncertainty by probability theory is limited.Although the traditional probability theory is the best way to deal with random uncertainty,it is not effective in dealing with cognitive uncertainty.Thus,DS evidence theory was proposed to study,which assigns probability to sets rather than disjoint single-element sets.As a reasoning method of uncertain information,it can reasonably describe incomplete,inaccurate and uncertain information,and reasoning uncertainties by calculating evidence to better support a hypothesis.So far,there are still some problems hindering the further development of evidence theory.For example,when there are many focal elements,there will be a lot of computation in the combination process,So many scholars have studied a lot of evidence of focal elements in the approximation process of belief function.On the basis of previous studies,this paper proposes a belief function approximation method which takes into account both the size of focus element cardinality and the size of belief value,and gives a standard TEC for evaluating approximation.Experiments show that this method can eliminate redundant information better and improve the effectiveness of evidence.In the actual application process,evidence theory will also face various uncertain conflict information.When evidence is synthesized,the greater the conflict between the evidence,the more often the synthesizing method can not get the result.Even if the result is obtained,it is often quite different from the fact.Therefore,many scholars at home and abroad have done a lot of research on how to deal with conflict evidence.On the basis of previous studies,this paper summarizes the advantages and disadvantages of existing fusion methods,and carries out in-depth research on these advantages and disadvantages.The research is mainly divided into two aspects.On the one hand,the research mainly focuses on how to modify the sources of evidence.On the basis of these studies,we propose a rank similarity index considering the order of focal elements of evidence,and give some properties of the rank similarity,and then use this index to modify the sources of evidence.On the other hand,it is found that when there is too much fused evidence,it is prone to noise and negative effects.Therefore,it is proposed to reduce the number of evidence by clustering the revised evidence in a cohesive hierarchical way,so as to weaken the influence of noise andnegative information.For clustered evidence,DS rule synthesis is used as the evidence representative of this kind of evidence,and reliability index of each class is used as the weight between classes.Finally,experiments show the effectiveness of the method.