Résumé:
Analytical queries defined on a star schema modeled data warehouse are very complex and time
consuming due to the join operations between the fact and dimension tables. Several techniques to reduce
the cost and response time has been emerged in the past decades such as indexes. Binary Join Indexes
(BJI) are one of the well-known indexes and its selection is considered as a problem itself (noted Index
Selection Problem: ISP). this problem is crucial in data warehousing physical design. To solve this
problem two approaches exists statistics-based approach and metaheuristic-based approach. In this
dissertation we propose a new metaheuristic-basedapproach. This approach is based on the improved
version of the artificial fish swarm algorithm for solving the binary join index selection problem. This
approach aims to select the optimal set of BJI based on a mathematical cost model. This method was
tested against a datamining constraint-based method and proved its effectiveness and even its superiority
to the datamining constraint method.