Research And Application Of Conflict Measurement In Complex Evidence Theory

In rapidly developing information age,intricate information and data fill people’s lives.However,how to integrate and judge the acquired data to achieve satisfactory results requires the concept of "information fusion".Information fusion refers to the process of combining different data,information,or knowledge from multiple sources into a more comprehensive and accurate information.However,incorrect decision-making in information fusion may have a negative impact on certain fields,such as national security,healthcare,and finance.In information fusion,conflict measurement is a crucial technology.When different sources of information exist in conflict,how to combine these pieces of information becomes a key problem.Conflict measurement refers to the technology of measuring the degree of conflict between different sources of information through a certain calculation method.Through conflict measurement,the similarity or difference between different pieces of information can be obtained,thus evaluating their consistency or contradiction.Therefore,the focus of this study is on how to measure the conflict between evidence sources and how to use conflict measurement methods to improve the accuracy of information fusion.In the theory of information fusion,the Dempster-Shafer(D-S)evidence theory has been widely applied due to its excellent ability to model uncertain information.As an extension of the D-S evidence theory,the complex evidence theory not only inherits the excellent properties and reliability of the D-S evidence theory,but also can perform information fusion in the environment of complex data characteristics and processing complex field data.Complex evidence theory,as a generalized uncertainty reasoning method,provides a unified uncertain information representation and processing platform for complex number fields.Therefore,in this thesis,we conduct research on conflict measurement methods and information fusion among evidence sources within the framework of the complex evidence theory.The specific research contents and methods are as follows:(1)Research on Conflict Measurement Based on Complex Evidence TheoryFor the issue of strictly measuring conflicts,this thesis proposes two improved conflict measurement methods.Firstly,we propose a new conflict measurement distance ICED,based on the classical conflict measurement method complex evidence distance(CED).Compared with CED,ICED improves the correlation between propositions.We verify the reliability of ICED by experimental analysis of complex basic belief assignment(CBBA)in multiple simulated complex and changing environments.Secondly,we propose a new conflict measurement method based on complex evidence theory.Complex belief function and complex plausibility function represent the total belief value and non-negative belief value of propositions,respectively.And these features can well express the differences between CBBAs.Based on this,a new distance measurement method CBD is proposed.The CBD can meet the requirements of distance properties,and through case studies and comparative experiments,this method has better performance than the previously proposed methods.(2)Research on the Application of Conflict Metric in Complex Evidence TheoryThe conflict measurement method plays a significant role in information fusion.To address the issue of how to use conflict measurement methods to improve the accuracy of information fusion,this thesis proposed a multi-source information fusion method based on the CBD.This method uses CBD as an effective conflict measure to compare and analyze different sources of evidence,and then integrates the results.This thesis also propose a decision fusion framework that can be used to optimize complex data,feature extraction and information fusion,thereby achieving the decision results.The effectiveness of the decisionmaking framework is validated through multiple sets of real-world datasets.The thesis focuses on improving and proposing new methods for conflict measurement to address the challenges and issues currently faced in information fusion.A decision fusion model is proposed to address the performance issues of information fusion using conflict measurement methods,based on the CBD method.Through numerous case studies and experiments using real-world data,the effectiveness of the proposed decision fusion model is demonstrated.Moreover,its potential for applications in medical diagnosis,predictive learning,and other fields is also established.