Personnalisation d'analyses décisionnelles sur des données multidimensionnelles

This thesis investigates OLAP analysis personalization within multidimensional databases. OLAP analyse is modeled through a graph where nodes represent the analysis contexts and graph edges represent the user operations. The analysis context regroups the user query as well as result. It is well described by a specific tree structure that is independent on the visualization structures of data and query languages. We provided a model for user preferences on the multidimensional schema and values. Each preference is associated with a specific analysis context. Based on previous models, we proposed a generic framework that includes two personalization processes. First process, denoted query personalization, aims to enhancing user query with related preferences in order to produce a new one that generates a personalized result. Second personalization process is query recommendation that allows helping user throughout the OLAP data exploration phase. Our recommendation framework supports three recommendation scenarios, i.e., assisting user in query composition, suggesting the forthcoming query, and suggesting alternative queries. Recommendations are built progressively basing on user preferences. In order to implement our framework, we developed a prototype system that supports query personalization and query recommendation processes. We present experimental results showing the efficiency and the effectiveness of our approaches.