Projective Clustering Ensembles

Recent advances in data clustering have regarded clustering ensembles and projective clustering methods, which distinctly aim to face typical issues in many clustering problems. In this paper, we address for the first time the projective clustering ensembles (PCE) problem, whose main goal is to derive a proper projective consensus partition from an ensemble of projective clustering solutions. We formalize PCE as an optimization problem which is designed to satisfy strong requirements on the independence on the specific clustering ensembles algorithm, ability to handle hard as well as soft data clustering, and different feature weightings. Specifically, we provide two formulations for PCE, namely a two-objective and a single-objective problem, in which the object-based and feature-based representations of the ensemble solutions are differently taken into account. Experiments have demonstrated the significance of the proposed methods for PCE, showing clear improvements in terms of accuracy of the output consensus partition.