The Credibility Decay Model Based On Evidence Theory

In recent years,the theory of evidence has attracted more and more attention because of its unique expression and fuzzy information processing.With the application of evidence theory in the field of information fusion,target recognition,decision-making,forecast and artificial intelligence is more and more widely,the relationship between the physical meaning of the related concepts of evidence theory and uncertainty theory has attracted some experts interest.At present,most research on information fusion are carried out in spatial domain,however,affected by the interference information and the performance of the sensor,the single sensor measurement period information is not accurate,but also the comprehensive utilization of identification information of a plurality of time nodes in time domain.Therefore,information fusion should be a process of time sequential multi sensor fusion based on.Although the temporal domain information fusion based on evidence theory has attracted the attention of some researchers,there is still a lack of temporal domain evidence combination method.The temporal evidence combination method needs to be further studied.It is necessary to construct the temporal evidence combination method based on the characteristics of temporal information fusion.It can be seen from the relevant research evidence theory,evidence combination is key to deal with the conflict of information,evidence of reliability assessment is an important method in dealing with the conflict,the key lies in the combination of time domain evidence how to deal with the conflict between the conflict of evidence.Therefore,this paper will focus on the study of dynamic reliability of sequential evidenceIn this paper,we first study the evidence fusion model,and make some reasonable improvements to the existing dynamic credibility decay model.Then,according to the characteristics of temporal fusion,the time decay model of evidence theory is applied to the analysis of human reliability,and the factors affecting the degree of dependence in human reliability analysis are analyzed.Finally,based on the generalized TOPSIS method,in the dynamic decision making,we make corresponding modification and apply the time decay model,which provides a scientific method for dynamic analysis of multi-attribute decision making.In order to demonstrate the effectiveness and scientific nature of these methods,we apply these algorithms to real data and compare them with the existing algorithms.The main work of this paper includes the following three aspects:1.According to the characteristics of temporal fusion,an improved credibility decay model is proposedIn the process of temporal fusion in the affected degree of reliability of the sensor,we modify the sensor to receive the information,these information will not only affected by the time,and may also be affected by the information content carried by self.Therefore,when we consider the credibility decay model,we add the influence factor of the information content,and integrate the difference of Shannon entropy.The experimental data and the traditional credibility decay model were compared,we proposed the decay model of information content based on time,tend to identify more and more scientific explanation in the time domain fusion process and the influence factors of credibility decay.2.Based on the time decay model,a new dependence assessment model is proposedIn the process of human reliability analysis,the most important one is to analyze the dependence of the event.There are three factors that directly affect the degree of dependence.One is the ”clossness in time”,the time decay model,expression of more scientific analysis to the influence factor,the specific time,digital,rather than the traditional fuzzy concept of time.Finally,we use specific experimental data to analyze the impact of these three factors,the degree of dependence on the impact.We use the evidence theory to fuse these three direct factors,and get the expression of the dependence degree of the event.We use the method of control variable to analyze the influence of a certain influence factor on the dependence,and the data value of the other two factors remain unchanged.Finally,we analyze the rationality of the experimental results.3.Based on the characteristics of the credibility decay model,we propose a dynamic generalized multi-attribute decision making methodIn the dynamic multi-attribute decision making,we give the historical data of the time node,the closer the time we are,the more help for us.This is the same as the idea of the credibility decay model,but the form is just the opposite,so we make some modifications to the credlibity decay model to meet the characteristics of dynamic multi-attribute decision making.Multi-attribute decision making is generalized to multi-attribute decision making compared to the traditional,classical multi-attribute decision making,request attribute evaluation model and framework of rules that must be the same,while the generalized decision does not require the same framework,to a greater degree of freedom expert.In this paper,we deal with the problem of dynamic multi-attribute decision making based on the improved credibility decay model and generalized multi-attribute decision making.Finally,we use the data of stock decision to illustrate the rationality and scientificity of our proposed method.