Algorithm 1: The Rule Generation Algorithm

Definition: Let k-itemSet denote an itemSet that contains k items. Let Pk denote the set of k-itemSet that may be frequent. Let Ck denote the set of k-itemSet that has a class label-derived item, whose sup and conf are denoted as supCount and confValue respectively. Let Fk denote the set of frequent candidate rules. Let CARk denote the set of CARs that have k items.

Input: The user-specified support threshold (minSup) and confidence threshold (minConf). All the frequent 1-items.

Output: The set of the extracted CARs, which is denoted as RuleSet.

1,   F1={frequent 1-items}

2,   for(k=2; Fk-1≠Ø; k++) do

3,     Pk=link(Fk-1);

4,     Ck={p∈ Pk∣p has a class label-derived item};

5,     Fk={c∈Ck∣c.supCount ≥ minConf}

6,     CARk={f∈Fk ∣ f.confValue ≥ minConf }

7,   end

8,   RuleSet=∪kCARk