Statistical preference as a comparison method of two imprecise fitness values

Abstract

The tness value of a knowledge base (KB) can be unknown and only some imprecise in- formation about it can be obtained. In some cases this information is given by means of an interval where we know the tness is con- tained. Thus, the comparison of two ran- domly distributed intervals is necessary in this context in order to be able to determine the preferences among individuals. This contribution is a rst approach to the use of statistical preference as a tool to compare this kind of intervals. We consider the probabilistic relation associated to the stochastic comparison of every pair of in- tervals and we study the cycle-transitivity of this relation. The defuzzi cation of this probabilitic relation, that is, the statistical preference relation, is studied and some properties are obtained. Our studies are particulary detailed for the case of the uniform distribution