Quantification of conflict in Dempster-Shafer framework: A new approach

被引:94
作者
George, T
Pal, NR
机构
[1] Machine Intelligence Unit, Indian Statistical Institute, Calcutta 700 035, 203, B.T. Road
关键词
conflict measure;
D O I
10.1080/03081079608945130
中图分类号
TP301 [理论、方法];
学科分类号
081202 ;
摘要
Uncertainty in Dempster-Shafer framework can be considered to consist of two components; one arising due to randomness, and the other arising due to the lack of specificity in evidence. One facet of uncertainty due to random: ness is conflict in the evidence. Here we establish the need for a new measure of conflict and follow a fresh approach to achieve the same. We view the conflict between propositions as being proportional to the 'distance'' between them. We expect this ''distance'' to obey the laws of a metric. With this motivation we formulate a set of axioms that we expect a metric distance to satisfy, in order to quantify the conflict between two propositions. Based on this we derive a unique expression for the conflict between propositions. The average of conflict between propositions gives a measure of fetal conflict in the body of evidence. We then prove various properties of this conflict measure.
引用
收藏
页码:407 / 423
页数:17
相关论文
共 14 条