Decision systems deal with a large volume of data stored in new databases called data warehouses. Data warehouses are typically modeled by a star schema that conventionally presents a central fact table and a set of dimension tables. The corresponding queries for this type of model are therefore very complex. In order to reduce the cost of executing complex queries, which contain very expensive joins, the solution envisaged would be to guarantee a good physical design of the data warehouses. Binary join indexes are very suitable to reduce the cost of executing these joins. In this work, we proposed a binary join index selection approach based on the notion of minimal transversal. The final configuration obtained is composed of several indexes, which make it possible to optimize the execution cost of the query set.