Dealing with numerical imprecision in mathematical programs for Electre Tri models disaggregation

Olivier Cailloux (1)

1 LGI - Laboratoire Génie Industriel - EA 2606

In this paper, we are interested in decision problems formulated as multicriteria sorting problems, i.e., when a finite number of alternatives evaluated on a set of criteria are to be assigned to one of an ordered set of predefined categories. The assignment is done based on the comparison of the alternatives to external norms, rather than by comparison of the alternatives to each other. Most of the work on preference elicitation in Multicriteria Decision Aid focuses on representing the preferences of a single decision maker (DM). We are interested in elicitation procedures for multiple DMs that make it possible for each DM to provide individual preference information in order to build a multiple criteria sorting model accepted by each DM as representing the group preferences. Three linear and mixed integer programs have recently been proposed to help solving the described problem. This technical report gives detailed technical informations about how the disaggregation programs proposed in the companion article have been implemented, including choice of constants and strategies to avoid numerical errors.

Domaines: Recherche opérationnelle [math.OC]

Olivier Cailloux. Dealing with numerical imprecision in mathematical programs for Electre Tri models disaggregation. 2014. ⟨hal-00985881⟩