The significance of a rule is the result of a statistical test to assess the likelihood that the strength of the rule is actually higher than that of its generalizations. A Fisher exact test is used to compare the rule’s strength to that of every rule that can be formed by deleting a single condition from the Left-Hand-Side of the rule. A further Fisher Exact Test is used to compare the strength of the rule to the frequency with which the Right-Hand-Side occurs in the data as a whole. The largest p value is reported. A small significance value indicates the rule probably represents an improvement on all of its generalizations.
The significance of an itemset is the smallest p value for tests for improvement relative to each partition of the itemset. A small significance value indicates that the itemset probably represents a real improvement upon all of its subsets.