Title: Resampling methods for comparing clustering solutions
Authors: Pietro Coretto - University of Salerno (Italy) [presenting]
Luca Coraggio - University of Naples Federico II (Italy)
Abstract: Selecting an optimal clustering solution is a longstanding problem. In model-based clustering this amounts to choose the architecture of the model mixture distribution. The main decisions to be made are: the cluster prototype distribution, the number of mixture components, and often other restrictions on the clusters' geometry. Classical penalized model selection criteria based on the observed likelihood function have been proposed to address this issue. We propose a selection strategy based on resampling, and we compare it with classical methods.