Dynamic Relation-Based Analysis of Objective Interestingness Measures in Association Rules Mining

While a large number of objective interestingness measures have been proposed to extract interesting rules from a dataset, most of them have been tested on a limited number of datasets that may not cover all possible patterns. This paper presents a framework to investigate relation among twenty-one interestingness measures on synthesized patterns (A → B), using all combinations of the six probabilities P(A, B), P(A, ¬B), P(¬A, B), P(¬A, ¬B), P(A) and, P(B) with a fixed number of occurrences. The partial order of interestingness measures is compared to that of another measure in order to characterize their similarity. The result shows 75 interrelation patterns of probabilities. An association rule mining is used to analyzed to describe for understanding their common and distinct properties.