Kemeny rule for preference aggregation: Reducing all exact solutions to a single one

Abstract

Preference aggregation as a single consensus ranking (CR) determination, using Kemeny rule, for an input profile consisting of m rankings, including ties, of n alternatives is used in multidimensional ordinal scale measurements, interval fusion with preference aggregation (IF&PA) technique, and various machine learning applications. In spite of its NP-hardness, the Kemeny rule can be fruitfully applied at the problem size n < 20, using branch and bound method, however, it can lead not to a single, but N optimal (exact) solutions (CRs), which are strict orders of the alternatives and form an output profile (OP). An efficient formal rule based on specific parameters of a weighted tournament matrix, constructed for the OP, is proposed to convolute the OP into an exact single final CR, which can include ties. The convolution rule is validated using Borda count, applied to the OP, and the notion of “betweenness” relation over rankings.