Addressing Overfitting and Underfitting in Gaussian Model-Based Clustering

UBC Faculty Research and Publications

Addressing Overfitting and Underfitting in Gaussian Model-Based Clustering Andrews, Jeffrey L.

Abstract

The expectation-maximization (EM) algorithm is a common approach for parameter estimation in the context of cluster analysis using finite mixture models. This approach suffers from the well-known issue of convergence to local maxima, but also the less obvious problem of overfitting. These combined, and competing, concerns are illustrated through simulation and then addressed by introducing an algorithm that augments the traditional EM with the nonparametric bootstrap. Further simulations and applications to real data lend support for the usage of this bootstrap augmented EM-style algorithm to avoid both overfitting and local maxima.

Item Metadata

Title	Addressing Overfitting and Underfitting in Gaussian Model-Based Clustering
Creator	Andrews, Jeffrey L.
Date Issued	2018
Description	The expectation-maximization (EM) algorithm is a common approach for parameter estimation in the context of cluster analysis using finite mixture models. This approach suffers from the well-known issue of convergence to local maxima, but also the less obvious problem of overfitting. These combined, and competing, concerns are illustrated through simulation and then addressed by introducing an algorithm that augments the traditional EM with the nonparametric bootstrap. Further simulations and applications to real data lend support for the usage of this bootstrap augmented EM-style algorithm to avoid both overfitting and local maxima.
Subject	EM algorithm; Bootstrap; Cluster analysis; Mixture models
Genre	Article; Postprint
Type	Text
Language	eng
Date Available	2025-03-05
Provider	Vancouver : University of British Columbia Library
Rights	Attribution-NonCommercial-NoDerivatives 4.0 International
DOI	10.14288/1.0448166
URI	http://hdl.handle.net/2429/90459
Affiliation	Science, Irving K. Barber Faculty of (Okanagan); Computer Science, Mathematics, Physics and Statistics, Department of (Okanagan)
Citation	Andrews, Jeffrey L. Addressing overfitting and underfitting in Gaussian model-based clustering, Computational Statistics & Data Analysis, Volume 127, 2018, Pages 160-171
Publisher DOI	10.1016/j.csda.2018.05.015
Peer Review Status	Reviewed
Scholarly Level	Faculty
Rights URI	http://creativecommons.org/licenses/by-nc-nd/4.0/
Aggregated Source Repository	DSpace

Open Collections

UBC Faculty Research and Publications