Multicriteria sorting using a valued preference closeness relation

To define a metric or a similarity relation with good properties is a crucial issue in most analogy and rule induction oriented methods for multicriteria sorting. Here, we propose a valued indifference (closeness) relation which is inspired on concordance and discordance measures. The criterion weights are obtained from the preferential information embedded in a reference set. This proposal performs very well in several practical problems. (C) 2007 Elsevier B.V. All rights reserved.