○ unseen · kind algorithm · level 0 · 0h
- Implementa: Unsupervised Learning
Mines “if-then” rules from transaction data — market-basket’s classic {bread, butter} → {milk}, scored by how often it holds and how much it actually means.
Mecanismo. Three scores per rule: support (how often the itemset appears at all), confidence (P(consequent | antecedent)), lift (how much more likely the consequent is given the antecedent vs. baseline — lift = 1 means no real association, just coincidence). The Apriori algorithm prunes the exponential search space with the anti-monotone property: if an itemset is infrequent, every superset of it is infrequent too — so frequent itemsets are built level-by-level (1-itemsets → 2-itemsets → …) instead of enumerating all 2^n subsets.
Ejercicio. h_comment_mining.R runs arules::apriori over co-occurring words/phrases in comments (wmjq_result_29/arules.txt, arulesViz.png) — real output to read against the mechanism above.
Enlaces
- Implementa: Unsupervised Learning