Entrepôts et analyse en ligne de données complexes centrés utilisateur : un nouveau défi

Abstract : The main goal of data warehouses is to facilitate decision making. In order to satisfy the whole analysis needs of the majority of the users, a promising issue consists in integrating a personalization process for OLAP analysis by taking into account user's own knowledge, preferences, needs,... In other words, the objective is to provide a user-centric decision-making system. In this thesis, we aim at proposing novel solutions for user-centric data warehouses. First, we have designed an original approach to achieve a user-driven model evolution that provides answers to personalized analysis needs. Our key idea consists in generating new anlysis axes based on users' knowledge by dynamically extending dimension hierachies or creating new ones. Moreover, to help users to find non expected and pertinent aggregates expressing deep relations within a data warehouse, we propose to combine data mining techniques with OLAP. We have more precisely defined a new roll-up operator based on the K-means clustering method. In addition, we have proposed a framework for mining large databases without size limit in very acceptable processing times. For this end, we have integrated data mining techniques within database management systems (DBMSs) by exploiting only their features. This helps to facilitate the extension of the capabilities of OLAP towards explicative and predictive analysis. To take into account both data sources changes and users requirements evolution, we have designed a user-centric approach for producing OLAP data cubes on the fly. This is based on a mediation system using ontologies. To generate the global merged ontology from local ontologies, we use the agglomerative hierarchical clustering method. Int the other hand, to warehouse complex data, we have designed a complex object-based multidimensional model. This is defined at two layers: (1) the package diagram layer which describes complex objects and their complex relationships and (2) the class diagram layer which provides details about the structure of each complex object. From the complex object-based multidimensional model, personalized complex object cubes can be derived. Eventually, for evaluationg our user-centric data warehouses solutions, we have implemented and carried out some experiments in both the contexts of relational and XML data warehouses.
Document type :
Habilitation à diriger des recherches
Complete list of metadatas

Cited literature [196 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00752126
Contributor : Fadila Bentayeb <>
Submitted on : Wednesday, November 14, 2012 - 7:25:58 PM
Last modification on : Thursday, November 22, 2018 - 11:41:17 AM
Long-term archiving on : Friday, February 15, 2013 - 6:05:11 AM

Identifiers

  • HAL Id : tel-00752126, version 1

Collections

Citation

Fadila Bentayeb. Entrepôts et analyse en ligne de données complexes centrés utilisateur : un nouveau défi. Base de données [cs.DB]. Université Lumière - Lyon II, 2011. ⟨tel-00752126⟩

Share

Metrics

Record views

1005

Files downloads

6697