Probabilistic Measures for Interestingness of Deviations - A Survey

Association rule mining has long being plagued with the problemof finding meaningful, actionableknowledge from the large set of rules. In this age of data deluge with modern computing capabilities, wegather, distribute, and store informationin vast amounts from diverse data sources.With suchdataprofusion, the core knowledge discovery problembecomes efficient data retrieval rather than simplyfindingheaps ofinformation. The most common approach is to employ measures of rule interestingness tofilter the results of the association rule generation process. However, study of literature suggests thatinterestingness is difficultto define quantitatively and can bebest summarized as, a record or pattern isinteresting if it suggests a change in an established model.Almost twenty years ago, Gregory Piatetsky-Shapiro and Christopher J. Matheus, in their paper, “TheInterestingness of Deviations,” argued that deviations should be grouped together in a finding and that theinterestingness of a finding is the estimated benefit from a possible action connected to it. Since then, thisfield has progressed and new data mining techniques have been introduced to address the subjective,objective, and semantic interestingness measures. In this brief survey, we review the current state ofliterature around interestingness of deviations, i.e. outliers with specific interest around probabilisticmeasures using Bayesian belief networks