Please use this identifier to cite or link to this item:
http://ir.lib.seu.ac.lk/handle/123456789/6314
Title: | Determining the optimal number of clusters using distance based k-means algorithm |
Authors: | Alibuhtto, M. C. |
Keywords: | Huge data Digital era Distance measure K-means algorithm |
Issue Date: | 15-Nov-2022 |
Publisher: | Faculty of Applied Sciences, South Eastern University of Sri Lanka, Sammanthura |
Citation: | Proceedings of the 11th Annual Science Research Sessions, FAS, SEUSL, Sri Lanka 15th November 2022 Scientific Engagement for Sustainable Futuristic Innovations pp. 60. |
Abstract: | In the current digital era, data is generated enormously with fast growth from different sources, and managing such huge data is a big challenge. Clustering algorithm is able to find hidden patterns and extract useful information from huge datasets. Among the clustering techniques, k-means clustering algorithm is the most commonly used unsupervised classification technique to determine the optimal number of clusters (k). However, the choice of the optimal number of clusters (k) is a prominent problem in the process of the k-means clustering algorithm. In most cases, clustering huge data, k is pre-determined by researcher, and incorrectly chosen k leads to increase computational cost. In order to obtain the optimal number of clusters, a distance-based k-mean algorithm was proposed with a simulated dataset. In the k-means algorithm, two distance measures were considered namely Euclidean and Manhattan distances. The results based on simulated data reveal that the k-means algorithm with Euclidean distance yields the optimal number of clusters compared to Manhattan distance. Testing on real datasets shows consistent results as the simulated ones. |
URI: | http://ir.lib.seu.ac.lk/handle/123456789/6314 |
ISBN: | 978-624-5736-60-7 978-624-5736-59-1 |
Appears in Collections: | 11th Annual Science Research Session - FAS |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
Fas symposium paper-25.pdf | 472.59 kB | Adobe PDF | ![]() View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.