Example. Assume that we have a relation r with
10 million rows over attributes A,. . . , F, and that we
want to find the frequent sets with the threshold 2 %.
Algorithm 1 randomly picks a small fraction s of r, say
20,000 rows, and keeps this sample s in main memory.
The algorithm can now, without any further database
activity, discover very efficiently sets that are frequent
in the sample.