Frequent Itemset Discovery.

摘要: Consider the set of all products sold by a supermarket. Assume that owner supermarker is interested in finding out subsets are often purchased together. Each customer transaction stored database, indicating The database can be described as table, whose columns (items), and rows transactions. value specific entry, is, (row, column)-pair, table 1 if corresponding product was transaction, 0 otherwise. task to find itemsets such items frequently occur same row (products together). most important interestingness measure frequent itemset mining support an itemset. It defined fraction contain x∈X. An its exceeds user-specified threshold value. Association rules closely related pattern class. Let R products, r X, Y ⊆ itemsets. Then X→ association rule over r. usually measured support(X→ Y)= support(X ∪ Y), confidence: conf (X → Y) = support(X∪Y,r) support(X,r) . Thus, confidence conditional probability randomly chosen from X also Given thresholds for confidence, given rules, supports confidences exceed thresholds.