Concise Representation of Frequent Patterns Based on Generalized Disjunction-Free Generators

Abstract

Frequent patterns are often used for solving data mining problems. They are applied e.g. in discovery of association rules, episode rules, sequential patterns and clusters. Nevertheless, the number of frequent itemsets is usually huge. In the paper, we overview briefly four lossless representations of frequent itemsets proposed recently and offer a new lossless one that is based on generalized disjunction-free generators. We prove on theoretical basis that the new representation is more concise than three of four preceding representations. In practice it is much more concise than the fourth representation too. An algorithm determining the new representation is proposed.