The use of structures and techniques of relational data warehouse is a subject of redress for the optimization of these. It helps to reduce the response time caused by the large volumes of fact tables, as well as the multitude of joints between it and the dimension tables.

Through our research, we proposed a new method based on ants. It consists to determine all the subsets of rows in the fact table that introduce the same predicates. These subsets are building the basis for partitioning the fact table.

Our work has been the subject of an experiment on the benchmark...