Determining OWA operator weights from ordinal relation on criteria

Abstract

The centered OWA (C-OWA) operator weights, defined as the center point of a convex hull constructed by the extreme points of ordinal relation, are computed by coordinate-wise averaging of the extreme points. The resulting C-OWA operator weights display some interesting properties. First there exists one to one correspondence between the multiplier used in the ordinal relation and the attitudinal character, thus we can confer a semantic to the attitudinal character in a meaningful way. Second, the C-OWA operator weights show a high level of coincidence with the maximum entropy OWA operator weights, which is widely used in OWA aggregation- another way of generating the OWA operator weights. This approach may help the decision-maker to articulate his or her preferences in lieu of direct assessment of attitudinal character. Indirectly assessed preference information will be used to generate the C-OWA operator weights, which are then used to aggregate multiple objects and finally prioritize multitude of alternatives.