We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
ZAGADNIENIE DOBORU LICZBY KLAS W KLASYFIKACJI SPEKTRALNEJ.
- Authors
Walesiak, Marek
- Abstract
The paper tested the usefulness of five indices assessing the quality of classification (within-group dispersion, Davies-Bouldin index, Caliński & Harabasz index, Hartigan index, Krzanowski & Lai index) in the issue of selection of the number of clusters in the spectral clustering taking into account four types of distance (squared Euclidean distance, Euclidean distance, Manhattan distance, GDM1 distance). The article evaluates twenty clustering procedures (four spectral clustering methods and five indices) based on two types of simulated data (classic and non-classic). Each clustering result was compared with the known cluster structure applying corrected Rand index.
- Subjects
AUTOMATIC classification; SPECTRAL theory; INDEXES; INDUSTRIAL clusters; RAND index; EUCLIDEAN distance
- Publication
Research Papers of the Wroclaw University of Economics / Prace Naukowe Uniwersytetu Ekonomicznego we Wroclawiu, 2013, Issue 278, p33
- ISSN
1899-3192
- Publication type
Article