Learning non-monotonic additive value functions for multicriteria decision making

Multiattribute additive value functions constitute an important class of models for multicriteria decision making. Such models are often used to rank a set of alternatives or to classify them into pre-defined groups. Preference disaggregation techniques have been used to construct additive value models using linear programming techniques based on the assumption of monotonic preferences. This paper presents a methodology to construct non-monotonic value function models, using an evolutionary optimization approach. The methodology is implemented for the construction of multicriteria models that can be used to classify the alternatives in pre-defined groups, with an application to credit rating.