The purpose of this study is to propose an ground motion selection algorithm for matching a design spectrum. Previous studies have shown that their own proposed algorithm required tedious and inefficient computation for scaling ground motions. Especially, a ground motion selection method of Kottke and Rathje (2008) revealed that limitations of using the algorithm became more pronounced with increasing the number of ground motions in a library because of insufficient computer memory and subsequent increase in computation time geometrically. The proposed algorithm achieved a good result that mean spectrum of selected ground motions could match up with a design response spectrum and efficient computation for scaling of ground motions.