A New Concise Representation of Frequent Patterns through Disjunctive Search Space

The interest in a further pruning of the set of frequent patterns that can be drawn from real-life datasets is growing up. In fact, it is a quite survival reflex towards providing a manageably-sized and reliable knowledge. This fact is witnessed by the proliferation of what is called concise representation of frequent patterns. In this paper, we propose an exact concise representation that explores the disjunctive search space in addition to the conjunctive one, in contrast with almost all known concise representations which only focussed on the latter space. This representation required the definition of a new disjunctive closure operator. The latter operator partitions the search space into distinct disjunctive equivalence classes and, hence, makes possible to drastically reduce the number of handled patterns. Empirical evidences are presented about the relative size of the new representation w.r.t. those based on frequent closed, (closed) non-derivable and essential patterns, respectively.