Skip to Main content Skip to Navigation
New interface

Modèles graphiques décomposables pour la décision individuelle et collective

Sergio Queiroz 1 
LIP6 - Laboratoire d'Informatique de Paris 6
Abstract : This thesis focuses on the use of GAI-Nets, a graphical model for compact preference representation, in order to achieve the usual features of a recommender system in the context where the space of alternatives has a large combinatorial size. Typically, web recommender systems use techniques well suited to highly standardized items such as CDs and DVDs, but inappropriate to combinatorial contexts. In addition, recommender systems for combinatorial contexts are often based on models that assume a kind of independence between attributes that ensures modeling preferences by an additive utility. GAI-Nets allow for interactions between attributes, thus being more general. Our key problems are choice and ranking of the k-best alternatives. We also study the problem of finding compromise solutions according to non-linear criteria in the context of multiperson/multicriteria decision making, as well as the problem of eliciting GAI-Nets. We provide adequate algorithms to solve these problems and, finally, we build a web application to apply the developed techniques in a concrete decision setting.
Complete list of metadata

Cited literature [128 references]  Display  Hide  Download
Contributor : Theses Bupmc Connect in order to contact the contributor
Submitted on : Friday, April 12, 2013 - 2:23:23 PM
Last modification on : Sunday, June 26, 2022 - 9:48:58 AM
Long-term archiving on: : Saturday, July 13, 2013 - 4:05:41 AM


  • HAL Id : tel-00812578, version 1


Sergio Queiroz. Modèles graphiques décomposables pour la décision individuelle et collective. Réseaux et télécommunications [cs.NI]. Université Pierre et Marie Curie - Paris VI, 2008. Français. ⟨NNT : 2008PA066357⟩. ⟨tel-00812578⟩



Record views


Files downloads