Constraint-based mining of formal concepts in transactional data

Abstract

We are designing new data mining techniques on boolean contexts to identify a priori interesting concepts, i.e., closed sets of objects (or transactions) and associated closed sets of attributes (or items). We propose a new algorithm D-Miner for mining concepts under constraints. We provide an experimental comparison with previous algorithms and an application to an original microarray dataset for which D-Miner is the only one that can mine all the concepts.