Résumé : Skyline paradigm provides an adequate tool to help users making decisions in the presence of multidimensional data and conflicting criteria. It is based on Pareto dominance relationship which returns the objects that are not dominated by any other object. The traditional skyline definition was recently expanded to group decision making to meet complex real-life needs encountered in many modern domain applications that require to analyze not only individual points but also groups of points. The group skyline aims at retrieving groups that are not dominated by any other group of the same size in the sense of a particular group-dominance relationship. It may often happen that this extended dominance leads to only a small number of non-dominated groups which could be insufficient for the decision maker. During the seminar, we expose our porposition to extend group skyline dominance by making it more demanding so that several groups leave incomparable. The original group skyline will be enlarged by some interesting groups that are not much dominated by any other group. The key element of this relaxation is a particular fuzzy preference relation, named « much preferred ». We also present some algorithms and some experimental results tested on real, synthetic and generated data to insure reliability and the decision quality of the proposition.
Sana Nadouri est ATER, Université de Bourgogne – ESIREM, Dijon, France.