Generalized disjunction-free representation of frequent patterns with negation
Marzena Kryszkiewicz · Journal of Experimental & Theoretical Artificial Intelligence · 2005
The discovery of frequent patterns is one of the main data mining problems. Frequent patterns are used for discovering association rules, episode rules, sequential patterns and clusters. The number of frequent patterns is usually huge. In order to alleviate this problem, a number of lossless representations of frequent patterns have been proposed recently. Theoretical works prove that frequent closed itemsets and models based on generalized disjunction-free sets or non-derivable itemsets are minimal representations of frequent patterns. Experiments prove further that the latter representations are by a few orders of magnitude more concise than the former one. The offered models were intended to represent only frequent patterns without negation. In this article, we offer a lossless representation based on generalized disjunction-free sets that represents frequent patterns both with and without negation. An algorithm for discovering this representation is also provided.