Résumé: In machine learning, the multiple classifiers aggregation problems consist in using multiple classifiers to enhance the quality of a single classifier. Simple classifiers as mean or majority rules are already used, but the aggregation methods used in voting theory or multi-criteria decision making should increase the quality of the obtained results. Meanwhile, these methods should lead to better interpretable results for a human decision-maker. We present here the results of a first experiment based on the use of Choquet integral, decisive sets and rough sets based methods on four different datasets.

---------

Ref HAL: hal-00660745_v1
Résumé: Ensemble methods (EMs) have become increasingly popular in data mining because of their efficiency. These methods(EMs) generate a set of classifiers using one or several machine learning algorithms (MLAs) and aggregate them into a single classifier (Meta-Classifier, MC). Amon MLAs, k-Nearest Neighbors (kNN) is one of the most known used in the context of EMs. However, handling the parameter k might be difficult. This drawback exists almost for all MLA that are instances based. Here, we propose an approach based on neighborhood graphs as alternative. Thanks to theses related graphs, like relative neighborhood graphs (RNGs) or Gabriel graphs (GGs), we provide a generalized approach with less arbitrary parameters. Introducing neighborhood graphs in EMs approaches has never been done before. The results of our algorithm : Neighborhood Random Classification are very promising since they are equal to the best EMs approaches such as Random Forest or those based on SVMs. In this exploratory and experimental work, we provide the methodological approach and we provide many comparison results.

Résumé: Utilisation des 4-capacités pour la modélisation dynamique de la prise de décision collective

---------

Ref HAL: hal-00625686_v1
Résumé: En apprentissage supervisé, les Méthodes Ensemble (ME) ont montré leurs qualités. L'une des méthodes de référence dans ce domaine est les Forêts Aléatoires (FA). Cette dernière repose sur des partitionnements de l'espace de représentation selon des frontières parallèles aux axes ou obliques. Les conséquences de cette façon de partitionner l'espace de représentation peuvent affecter la qualité de chaque prédicteur. Il nous a semblé que cette approche pouvait être améliorée si on se libérait de cette contrainte de manière à mieux coller à la structure topologique de l'ensemble d'apprentissage. Dans cet article, nous proposons une nouvelle ME basée sur des graphes de voisinage dont les performances, sur nos premières expérimentations, sont aussi bonnes que celles des FA.

Résumé: A group of agents is faced with collective decisional problems. The corresponding debate is seen as a dynamical process. A first theoretical model based upon a muticriteria decision framework was proposed in (Rico et al., 2004) but without semantic justifications and explicit dynamical representation. A second descriptive model was proposed in (Imoussaten et al., 2009) where social influences and argumentation strategy govern the dynamics of the debate. This paper aims at justifying the equations introduced in (Rico et al., 2004) with the semantics concepts reported in (Imoussaten et al., 2009) to provide a model of a debate in the framework of control theory that explicitly exhibits dynamical aspects and offers further perspectives for control purposes of the debate.