Abstract
dc:descriptionIn this paper, a comparison of the three recently published novel clustering methods: Parameter free Clustering based K means, A Criterion for Deciding the Number of Clusters in a Dataset and Graph-based Data clustering via Multiscale Community Detection is undertaken. Each of the these methods are replicated on both real and synthetic data sets for which their performance is evaluated with the regards to the relative error between the number of outputted clusters and the true cluster number for each data set. It was found that the Graph-based Data clustering via Multiscale Community Detection was the best performing and was able to determine the cluster number for each data set with complete accuracy with regards to the number of partitions. Further evaluation of each method considers the quality of the output in more depth as well as evaluates the strengths and limitations of each respective method.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Kilaire, Shanice,
Subjects
dc:subject × 1Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- eng
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD192057777&prefLang=en_US