Warehouses are obliged to optimize their operations with regard to multiple objectives, such as maximizing effective use space, equipment, labor, maximize accessibility of products, maximize amount of processed orders and all this should be achieved whilst minimizing order processing times, distance traveled, broken promises, errors and not to forget the operational cost. A product placement problem for a warehouse is in focus of this study and the main goal is to decrease the picking time for each pick run in order to gain higher efficiency. To achieve this, a simulation model is built as a representation of the warehouse. As the complexity and the size of the number of input variable grow it is essential to use simulation-based optimization in order to receive a satisfying result. A set of initial solutions for the simulation-based optimization is needed; since the number of products to place in the warehouse is huge this solution ought to be intelligent. This paper describes a technique for generating such a set of solutions through searching for frequent itemsets in the transaction database. It is believed that frequent products usually picked simultaneously should be stored closed together.