While faster and faster algorithms have been presented for finding frequent patterns, but, the memory issue is almost always neglected. To study this problem, this paper presents our heuristic that maps the input data set onto a smaller one. The essential idea is to determine the right choice of the tuning number by the reward measure introduced in this work. The tuning number helps us to find the maximal number of repeats in the input data set. Experiments done on several large data sets show the benefit of our method in data mining.