一种新的证据冲突计算方法及其应用

A New Calculation Method of Evidence Conflict and Its Applications

证据理论在处理不确定信息方面具有建模方便、算法收敛速度快等优势,但也存在不能有效处理高冲突信息的不足。针对Jousselme证据距离函数不能准确描述证据概率分配值较分散的证据间的冲突度量问题,本文提出了随证据概率分布之间非包含度的增大冲突度量结果按比例增大的改进冲突度量函数,并将其应用于解决实际应用中的风险概率预测、目标识别问题,同时与已有其他冲突度量算法进行对比分析,验证了所提算法的有效性和广泛适用性。

Evidence theory has the advantages of convenient modeling and fast algoritlim convergence in dealing with uncertain information,but it also has the disadvantage of not being able to deal effectively with high-conflict information.According to the problem that Jousselme evidence distance function can't accurately describe the conflict measure between pieces of evidence with more scattered probability distributions,an improved conflict measure function was proposed which increased the conflict measure proportionally with the increase of non-inclusive degree between evidence probability distributions,and applied to solving problems such as lisk probability prediction and target recognition in practical applications.The effectiveness and wide applicability of the proposed algorithm are verified in comparison with other existing conflict measure algorithms.