Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)


Mathematical Sciences

First Advisor

Nezamoddin Nezamoddini-Kachouie

Second Advisor

Veton Kepuska

Third Advisor

Jewgeni Dshalalow

Fourth Advisor

Munevver Subasi


Due to advancements in data acquisition, large amount of data are collected on a daily basis. Analysis of the collected data is an important task to discover the patterns, extract the features, and make informed decisions. A vital step in data analysis is dividing the subjects (elements, individuals) in different groups based on their similarities. One way to group the subjects is clustering. Clustering methods can be divided into two categories, linear and non-linear. K-means is a commonly used linear clustering method, while Kernel K-means is a non-linear technique. Kernel K-means projects the elements to a new space using a kernel function and then clusters them in different groups. Different kernels perform differently when they are applied to different data sets and as a result choosing the right kernel for an application could be challenging. Therefore, applying a set of kernels and aggregating the results could provide a robust performance for different data sets. In this work, we address this issue and propose a weighted majority voting to ensemble the results obtained by different kernels.

Included in

Mathematics Commons