摘要 |
A method and apparatus for mining generally long patterns from a database of data records of items. An initial set C of candidates is first generated, each candidate c having disjoint sets of items c.head and c.tail. Frequent candidates from the set C are extracted and put into a set F, where the frequent candidates are those whose set {c.head &org& c.tail} is an itemset having a minimum support. Non-frequent candidates in C are used to generate new candidates, which are added to the set C. After any candidates having a superset in the set Fare removed from C and F, the method steps are repeated on the new candidate set C, until C is empty. The candidates remaining in the working set Fare returned as the desired patterns.
|