Exploring interestingness through clustering: a framework

Abstract

Determining interestingness is a notoriously difficult problem: it is subjective and elusive to capture. It is also becoming an increasingly more important problem in knowledge discovery from database as the number of mined patterns increases. In this work we introduce and investigate a framework for association rule clustering that enables automating much of the laborious manual effort normally involved in the exploration and understanding of interestingness. Clustering is ideally suited for this task; it is the unsupervised organization of patterns into groups, so that patterns in the same group are more similar to each other than to patterns in other groups. We also define a data-driven inferred labeling of these clusters, the ancestor coverage, which provides an intuitive, concise representation of the clusters.