The Modification of the K-means Method for Creating Non-convex Clusters

Result code in IS VaVaI
<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F44555601%3A13510%2F14%3A43886630" target="_blank" >RIV/44555601:13510/14:43886630 - isvavai.cz</a>
Result on the web
—
DOI - Digital Object Identifier
—

Result language
angličtina
Original language name
The Modification of the K-means Method for Creating Non-convex Clusters
Original language description
Cluster analysis involves many different methods based on a variety of principles. Each of these methods has its advantages and disadvantages. The aim of authors of the new algorithms is to search for new methods that use existing methods positives and minimize the negatives. In this article, we discuss one of the most famous and most used method, the k-means method. The inability to search non-convex clusters is one of known weak points of the k-means method. Nevertheless, this method has many advantages such as simplicity and speed. This algorithm is also implemented in a lot of statistical software. The article presents a proposal for a solution. The second phase of the process is proposed. We create a certain number of clusters which is larger thandesired in the first stage of processing by the classical algorithm k-means. Then we combine some clusters using appropriate agglomerative methods and reduce the number of clusters required in the second phase. There is the proposed proc
Czech name
—
Czech description
—

Project
—
Continuities
I - Institucionalni podpora na dlouhodoby koncepcni rozvoj vyzkumne organizace

Publication year
2014
Confidentiality
S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Article name in the collection
8th International Days of Statistics and Economics
ISBN
978-80-87990-02-5
ISSN
—
e-ISSN
—
Number of pages
9
Pages from-to
1722-1730
Publisher name
Melandrium
Place of publication
Slaný
Event location
Praha
Event date
Sep 11, 2014
Type of event by nationality
EUR - Evropská akce
UT code for WoS article
000350226700169

Similar results(10)