一种修正证据距离

Jousselme提出证据距离衡量一组表示分类结果的证据与代表真实结果证据之间的差异性程度。该文通过分析得到Jousselme证据距离衡量一般形式证据的差异性程度存在不足：证据基本信任分配分布的分散程度越大，Jousselme证据距离越小，两组完全不相同的非类别证据间的Jousselme证据距离不为最大值1。针对不足，该文提出一种修正证据距离。数值算例表明，修正证据距离扩展了Jousselme证据距离对证据差异性程度衡量的适用范围。

A Modified Distance of Evidence

Jousselme advanced the distance of evidence to measure the degree of difference between evidence that represents the identification result and evidence that represents the real solution. The distance of evidence becomes smaller with the increase of dispersion of evidences basic probability assignment. And for two totally different evidences, the distance is less than one unless the two evidences are both category ones. Therefore, the distance of evidence proposed by Jousselme is unsuitable to measure the difference between two evidences. Aiming at this problem, a modified distance of evidence is put forward. Numerical examples indicate that the modified distance of evidence extendes the application scope of distance proposed by Jousselme for measuring the difference between evidences.