Un modèle multidimensionnel pour un processus d'analyse en ligne de résumés flous. (A multidimensional model for on-line analytical processing of fuzzy summaries)

The work presented in this thesis deals with the subject of exploration and manipulation of database summaries with significant size. The summaries produced by SaintEtiQ system are multilevel materialized views of homogeneous data clusters, presented with a collections of fuzzy labels available on each attribute. Our thesis contribution is based on three points. Initially we defined a logical data model called summaries partition, by analogy with OLAP datacubes, with the aim of offering to the end-user a tool for data presentation in condensed form and adapted to the analysis. Secondly, we defined a collection of algebraic operators on the multidimensional space of summaries partitions. These operators are the base for an algebra for handling summaries. This algebra takes into account specificities of the summary model we deal with. We adapted the majority of the operators of analysis proposed in OLAP systems. Thus, we identified: core operators resulting from the relational algebra, operators of changing granularity and operators of reorganization. These results offer new prospects for the effective summaries exploitation in a decisional system. Finally, to complet this work, we were interested in the summaries and partitions representation, in particular to provide a clear and concise presentation of it to the end-user. Applied to a summaries hierarchy produced by the SaintEtiQ system, the approach tries to construct fuzzy prototypes representing the summaries.