Combination Method of Conflict Evidences Based on Similarity Dynamic Corrected
Feng Zhou, Jin Gou, Jiliang Luo · Journal of Chinese Computer Systems · 2012
Two traditional improvement strategies,modified DS(Dempster-Shafer) combination rules and modified evidence model,are used to solve some typical combination problems of conflict evidences in information fusion subject,such as general conflict,vote-down,and robustness.A new method of modified model based on closed world assumption named CESDC(Conflict Evidence Similarity Dynamic Corrected) is proposed in this paper.Jousselme distance function is introduced to quantify the correlations between focal attributes and evidences,based on which the support degree value of each evidence could also be calculated.Referenced evidence mentioned in this paper is obtained by the weighted average of all evidences according to their support degree values.The referenced evidence is used for uncertainty verification to count out the value similarity and direction similarity between each evidence and referenced evidence.It also establishes a dynamic correction model of the value similarity and direction similarity,the corrected result of which uses the DS combination rules to achieve the evidences combination.Then we set the average value of those multi-group results,which are all computed by the dynamic correction model,as the final fusion result.Experiment results show the effective and reasonable of the proposed method.