Learning Characteristic Rules Relying on Quantified Paths

Teddy Turmeaux (1) , Ansaf Salleb (1) , Christel Vrain (1) , Daniel Cassard

1 LIFO - Laboratoire d'Informatique Fondamentale d'Orléans

In this paper, we address the characterization task and we present a general framework for the characterization of a target set of objects by means of their own properties, but also the properties of objects linked to them. According to the kinds of objects, various links can be considered. For instance, in the case of relational databases, associations are the straightforward links between pairs of tables. We propose , a new algorithm for mining characterization rules and we show how it can be used on multi-relational and spatial databases.

Mots clés: Machine Learning; Inductive Logic Programming; Data Mining; Characteristic Rules; Relational Databases; Spatial Databases

Domaines: Apprentissage [cs.LG]; Intelligence artificielle [cs.AI]

Teddy Turmeaux, Ansaf Salleb, Christel Vrain, Daniel Cassard. Learning Characteristic Rules Relying on Quantified Paths. 2003, pp.471-482, ⟨10.1007/b13634⟩. ⟨hal-00084885⟩