Many association analysis algorithms rely on an Apriori-like approach for finding frequent patterns. The overall structure of the algorithm is given below.

Suppose we are interested in finding boolean logical rules such as



{a ? b} ?? {c, d},



which may contain both disjunctions and conjunctions of items. The corre-

sponding itemset can be written as {(a ? b), c, d}.



(a) Does the Apriori principle still hold for such itemsets?

(b) How should the candidate generation step be modified to find such

patterns?

(c) How should the candidate pruning step be modified to find such pat-

terns?

(d) How should the support counting step be modified to find such pat-

terns?

Refer to R. Srikant, Q. Vu, R. Agrawal: Mining Association Rules
with Item Constraints. In Proc of the Third Int’l Conf on Knowledge
Discovery and Data Mining, 1997.