High-dimensional Clustering with Random Projections

Random projections (RPs) have shown to provide promising results for high-dimensional classification. In this work, we address the issue of high-dimensional clustering by exploiting the general idea of RP ensemble to perform unsupervised classification. Specifically, we generate a set of low dimensional independent random projections and we perform a model-based clustering on each of them. The top B1 projections, i.e. the ones showing the best grouping structure according to different cluster quality measures, are then selected. The final partition is obtained by aggregating, via consensus, the chosen classifiers. The performances of the method are assessed on a set of both real and simulated data.