1. Generate all frequent item-sets, i.e., all sets Z such that NZ /N ≥ minsup and
|Z|≥ 2.
For each frequent item-set Z, consider all partitionings of Z into two nonempty subsets X and Y . If confidence(X ⇒ Y) ≥ minconf , then keep the rule X ⇒ Y .
If confidence(X ⇒ Y) < minconf , then discard the rule.
2. Output the rules found.
This simple algorithm has two problems. First of all, there is a computational problem related to the first step. If there are m variables, then there are 2m − m – 1 possible item-sets. Hence, for 100 products (m = 100) there are