{Saket Choudhary}

Pattern Discovery

Pattern: Set of items/subsequences that occur frequently in a dataset. Intrinsic property of the dataset

  • Absoulte Support: Frequency of X in dataset
  • Relative Support: Absolute Support/ Total # of records

Itemsets to association rules

  • Association Rule: $X \Longrightarrow Y(S, c)$
  • Support(s): Probability that a transaction contains $X \cup Y$
  • Confidence(c): The conditional probability that a transaction containing X also contains Y: $c = \frac{Sup(X \cup Y)}{sup(X)}
  • Association rule mining: Find all the rules $X \Longrightarrow Y$ with minimum support and confidence

Closed Patterns

  • Too many frequent patterns(all possible subsets): $2^n-11$
  • Closed Patterns: X is closed if X is frequent(it satisfies the minsup threshold) and there exists no supper pattern $X \subset Y$ with same support as X