Assessing Rule Interestingness with a Probabilistic Measure of Deviation from Equilibrium
Julien Blanchard, Fabrice Guillet, H. Briand, Régis Gras · 2005
Assessing rule interestingness is the cornerstone of successful applications of association rule discovery. In this article, we present a new measure of interestingness named IPEE. It has the unique feature of combining the two following characteristics: first, it is based on a probabilistic model, and secondly, it measures the deviation from what we call equilibrium (maximum uncertainty of the consequent given that the antecedent is true). We study the properties of this new index and show in which cases it is more useful than a measure of deviation from independence.