In this paper we describe a computer intensive method to find the ridge parameter in a prediction oriented linear model. With the help of a factorial experimental design the method is tested and compared to a classical one.