A method for ordinal classification in multicriteria decision making

In a discrete multicriteria decision problem, a finite set of alternatives are evaluated in terms of multiple criteria. When the goal is to partition the alternatives into predefined ordered categories, this is called an ordinal classification or sorting problem. In this paper we present the new SMAA-OC method for the ordinal classification problem that can handle uncertain, imprecise or partially missing criteria and preference information. SMAA-OC is based on a utility or value function to represent the DMs' preference structure and boundary profiles to define the categories. The method is implemented through stochastic simulation and statistical analysis. We demonstrate the method with a small example.