Title: Random projection ensemble clustering
Authors: Francesca Fortunato - University of Bologna (Italy)
Laura Anderlucci - University of Bologna (Italy)
Angela Montanari - Alma mater studiorum-Universita di Bologna (Italy) [presenting]
Abstract: Random projections (RPs) have shown to provide promising results for high-dimensional classification. The RP ensemble classifier, in fact, overcomes the inherent instability of a single RP by using a ``selection and ensemble aggregation'' routine. The general idea of RP ensemble is extended to high-dimensional clustering purposes. Specifically, following the original procedure, the best projection according to a specific clustering quality measure is chosen within each of $B_1$ distinct blocks of $B_2$ RPs. Then, the final partition is obtained by aggregating, via consensus, results of applying a model-based clustering algorithm to the selected projections. The performances of the method are assessed both on a set of real and simulated data.