Term of Award
Master of Science in Applied Mathematics
Document Type and Release Option
Thesis (restricted to Georgia Southern)
Committee Member 1
Committee Member 2
This project will extend the family of clustering criterion
Dm,p(U, v1,v2,...,vc=ni=1Σ*ck=1ΣUmik II xk-viIIpp
to include the case of 0 < p < 1, and then to empirically examine how the choices of m and p in the extended family affect the influence of data outliers on cluster solutions. The study of this extended family of clustering methodology consists of the following three parts:
devising a reliable approach to optimizing Dm.p based on a combination of techniques such as alternating optimization, bisection, and exhaustive search;
implementing the optimization scheme in MATLAB;
conducting numerical experiments to learn how the choices of m and p affect the clustering results.
Overstreet, Dessa Darlene, "Generalized Fuzzy c-Means Clustering" (1998). Legacy ETDs. 85.