Title: Clustering via a semi-parametric density estimation
Authors: Mahdi Salehi - University of Neyshabur-University of Pretoria (Iran) [presenting]
Andriette Bekker - University of Pretoria (South Africa)
Mohammad Arashi - Ferdowsi University of Mashhad (Iran)
Abstract: The idea behind density-based clustering is to associate groups with the connected components of the level sets of the density of the data to be estimated by a nonparametric method. This approach claims some advantages over both distance- and model-based clustering. Some researchers developed this technique by proposing a graph theory-based method for identifying local modes of the underlying density being estimated by the well-known kernel density estimation (KDE) with normal and t kernels. The aim is to improve the performance of the density-based clustering by using a semi-parametric KDE with a more flexible family of kernels belonging to skew-symmetric distribution. Finding optimum bandwidth under the mentioned kernels is another main result where we shrink the bandwidth more than the one obtained under the normality assumption. Finally, some illustrative examples follow.