摘要
证据分类策略是一种有效的冲突证据推理融合方法,但是在实际应用中会产生大量重复分类,而且分类门限根据主观经验确定,缺乏依据。为了降低分类数,提高证据分类策略的准确性,提出一种基于证据相似性的证据分类策略。首先以证据推理中得到的初始证据作为理想证据,通过衡量各个证据与理想证据的证据距离,按照证据的相似性将系统内证据分为2类证据集,然后继续按照以上的分类策略分类,直到不可再分。最后对分类结果采用Dempster组合规则合成,衡量各个分类的可信度,对合成结果加权平均组合。通过算例对该策略进行了验证,结果表明:基于证据相似性的证据分类策略可以有效降低分类数。
The evidence classification strategy is a useful conflict evidence reasoning and fusion method, but in practice it would bring a great deal of repetitive class. In order to reduce the number of class and improve the efficiency of evidence classification strategy, an evidence classification strategy based on similarity of evidence is proposed. First by taking the initially evidence in evidence reasoning as the ideality evidence, scaling the distance between each evidence and the ideality evidence, the evidences in the system are divided into two classes according to the similarity of them. Then the classification will continue with the above classification strategy till it will be impossible. Finally the following workings such as combining the classified results with Dempster combination rule, scaling the reliability of the classification results and weighting the Dempster combination result are done. The strategy is validated through an example. The result shows that the evidence classification strategy based on similarity of evidence can reduce the number of class effectively.
出处
《空军工程大学学报（自然科学版）》
CSCD
北大核心
2009年第3期55-58,共4页
Journal of Air Force Engineering University(Natural Science Edition)
基金
国家自然科学基金资助项目(60773209)
关键词
证据理论
分类策略
证据相似性
evidence theory
classification strategy
similarity of evidence
