Association Rule Mining
Rules describing which items co-occur more than chance predicts, scored by support, confidence, and lift, the last of which rules out spurious associations.
Confidence alone can't distinguish a real association from an item that's simply popular overall, since a popular item shows up at a high rate no matter what triggered the rule. Lift divides that conditional rate by the item's own baseline popularity, and a lift of exactly one means no real association exists at all.
Apriori prunes the search using the fact that no itemset can be more frequent than any of its subsets; FP-Growth avoids repeated full-database scans by mining a compressed tree instead.