B0336
Title: Graphical tools for modal clustering diagnostics
Authors: Jose E Chacon - Universidad de Extremadura (Spain) [presenting]
Abstract: Modal clustering is a nonparametric methodology that uses density estimation to find clusters. The basic idea is to estimate the modes of the density, and then assign points to the modes by finding the basins of attraction of the modes. We present diagnostics for modal clustering. Particularly, we provide graphical tools to assess the uncertainty in the cluster assignment of a data point. As expected, the higher uncertainty zones are located around the cluster boundaries.
Journal