Bridging Conjunctive and Disjunctive Search Spaces for Mining a New Concise and Exact Representation of Correlated Patterns

In the literature, many works were interested in mining frequent patterns. Unfortunately, these patterns do not offer the whole information about the correlation rate amongst the items that constitute a given pattern since they are mainly interested in appearance frequency. In this situation, many correlation measures have been proposed in order to convey information on the dependencies within sets of items. In this work, we adopt the correlation measure

bond

, which provides several interesting properties. Motivated by the fact that the number of correlated patterns is often huge while many of them are redundant, we propose a new exact concise representation of frequent correlated patterns associated to this measure, through the definition of a new closure operator. The proposed representation allows not only to efficiently derive the correlation rate of a given pattern but also to exactly offer its conjunctive, disjunctive and negative supports. To prove the utility of our approach, we undertake an empirical study on several benchmark data sets that are commonly used within the data mining community.