3 Mining transactional data
Mining transactional data —frequent item-sets chapter 6 Frequent itemsets: sup ( I ) >= s Support of I: sup ( I ) = # baskets that contains all items in I Support ratio = sup ( I ) /n n = # baskets Maximally frequent: I is frequent , none of I ’ s supersets is frequent Applications: 1. Baskets—transactions// documents 2. Items—products// words 3. Frequent item sets: products frequently bought together// Words that occur together frequently in many documents Association rules: I —>j ( I: a set of items , J: an item ) I: a set of items , j: an item If all of items in I appear in some baskets , then j likely appears in that baskets too Confidence ( I->j ) = sup ( I U{j} ) /sup ( I ) Finding rules with high confidence: I —>j: both of sup ( I U{j} ) and sup ( I ) should be high For frequent item sets For each of them j , determine if J-{j}—>j has high confidence for j belongs t...