Thèse soutenue

Exploration collaborative de cubes de données

FR  |  
Auteur / Autrice : Elsa Negre
Direction : Arnaud GiacomettiPatrick Marcel
Type : Thèse de doctorat
Discipline(s) : Informatique
Date : Soutenance le 01/12/2009
Etablissement(s) : Tours
Ecole(s) doctorale(s) : Ecole doctorale Santé, sciences, technologies (Tours)
Partenaire(s) de recherche : Equipe de recherche : Laboratoire d'Informatique Fondamentale et Appliquée de Tours (2012-...)
Laboratoire : École polytechnique universitaire (Tours)
Jury : Président / Présidente : Mokrane Bouzeghoub
Examinateurs / Examinatrices : Georges Hebrail
Rapporteurs / Rapporteuses : Franck Ravat


FR  |  

Data warehouses store large volumes of consolidated and historized multidimensional data to be explored and analysed by various users. The data exploration is a process of searching relevant information in a dataset. In this thesis, the dataset to explore is a data cube which is an extract of the data warehouse that users query by launching sequences of OLAP (On-Line Analytical Processing) queries. However, this volume of information can be very large and diversified, it is thus necessary to help the user to face this problem by guiding him/her in his/her data cube exploration in order to find relevant information. The present work aims to propose recommendations, as OLAP queries, to a user querying a data cube. This proposal benefits from what the other users did during their previous explorations of the same data cube. We start by presenting an overview of the used framework and techniques in Information Retrieval, Web Usage Mining or e-commerce. Then, inspired by this framework, we present a state of the art on collaborative assistance for data exploration in (relationnal and multidimensional) databases. It enables us to release work axes in the context of multidimensional databases. Thereafter, we propose thus a generic framework to generate recommendations, generic in the sense that the three steps of the process are customizable. Thus, given a set of sequences of queries, corresponding to the previous explorations of various users, and given the sequence of queries of the current user, our framework proposes a set of queries as recommendations following his/her sequence. Then, various instantiations of our framework are proposed. Then, we present a Java prototype allowing a user to specify his/her current sequence of queries and it returns a set of recommendations. This prototype validates our approach and its effectiveness thanks to an experimentations collection. Finally, in order to improve this data cube exploration collaborative assistance and, in particular, to share, navigate or annotate the launched queries, we propose a framework to manage queries. Thus, an instantiation to manage recommendations is presented.