Feasible Optimization - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2016

Feasible Optimization

Résumé

We consider a decision problem related to the future configuration selection in a complex system (for example an enterprise) to improve its performances. We formulate the problem considering a set of criteria representing the axis on which the performances of the considered system are evaluated, and a set of alternative configurations of the system, possibly infinite. It means that a priori every configuring of the system can be considered. We consider that overall preference or indifference relation can be defined for a given pair of alternatives, each alternative being represented by a value function. We consider also a feasibility function such as it is possible to pass from a given alternative to another one if and only if the feasibility of the first alternative is higher than the second one. Also the feasibility function is non decreasing in all its arguments. In this context, the optimal configuration of the considered system, having current configuration could be obtained by the solving the following optimization problem: maximizing the utility value under the constraint of feasibility. We assume that both the value function and the feasibility function have an additive form. Moreover the both marginal utility and feasibility according each criterion are non decreasing. We propose to assess the value and feasibility functions in an indirect way by ordinal regression through the UTA method, which is a rather well known procedure for obtaining a value function.
Fichier non déposé

Dates et versions

hal-01379094 , version 1 (11-10-2016)

Identifiants

  • HAL Id : hal-01379094 , version 1

Citer

Vincent Clivillé, Corrente Salvatore, Bastien Rizzon, Salvatore Greco. Feasible Optimization. 27th European conference on operational research, Jul 2015, Glasgow, United Kingdom. ⟨hal-01379094⟩
32 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More