Theoretical bounds on the size of condensed representations
KDID'04 Proceedings of the Third international conference on Knowledge Discovery in Inductive Databases
Key roles of closed sets and minimal generators in concise representations of frequent patterns
Intelligent Data Analysis
| Hi-index | 0.00 |
A number of lossless representations of frequentpatterns were proposed in recent years. Therepresentation that consists of all frequent closeditemsets and the representations based on generalizeddisjunction-free patterns or on non-derivable itemsetsare proven the most concise ones. Experiments showfurther that the latter ones are by a few orders ofmagnitude more concise (and determinable) than theformer one. As follows from experiments, therepresentations based on generalized disjunction-freepatterns are also more concise than the available in theliterature representations of frequent patterns, whichdetermine supports of patterns in an approximate way. Inthis paper, we provide an upper bound on the length ofgeneralized disjunction-free patterns. The bounddetermines the maximum number of scans of thedatabase carried out by Apriori-like algorithmsdiscovering the representations based on generalizeddisjunction-free patterns.