Algoritmos evolucionários e técnicas de tomada de decisão em análise multicritério

Abstract

The multicriteria analysis studies manners of aiding man (in this context, called decision-maker) to make decisions in presence of uncertainty and conflicting interests. As a unique point-of-view is insufficient to embrace all necessary information and all contradictions inherent to the problem, generally, the decision analysis needs to handle many criteria simultaneously. Multiobjective optimization problems have a set of efficient solutions. As in real applications only one solution from this set is executed, a decision problem is originated: choose between many optimal alternatives the one that fits the decision-maker preferences, considering simultaneously several criteria. In this context, this thesis studies the multicriteria analysis and the use of multicriteria decision methods, associated to evolutionary optimization algorithms, in order to completely solve multiobjective optimization problems. This association takes advantage of the multicriteria analysis effectiveness in handling concurrent interests and of the evolutionary search efficiency in meeting optimal solutions. This work contains two main contributions: Two new decision methods: the Multiplicative Promethee, which is, in fact, a new version of the popular decision method, Promethee II, and the Multicriteria Tournament Method. A new interactive decision technique, where Promethee II is coupled to a mono-objective evolutionary search algorithm. Both decision methods produced satisfactory results when they were used in the choice of the final solution for multiobjective optimization problems. In addition to analytical problems with different characteristics, two electromagnetic problems, stated as optimization ones, were also considered in this study. The proposed interactive decision technique was also successful in meeting the efficient solution preferred by the decision-maker.