Welcome!
To use the personalized features of this site, please log in or register.
If you have forgotten your username or password, we can help.
|
 |
On Filtering the Training Prototypes in Nearest Neighbour Classification
| |
|
On Filtering the Training Prototypes in Nearest Neighbour Classification
J. S. Sánchez3, R. Barandela4 and F. J. Ferri5
| (3) |
Dept. Llenguatges i Sistemes Informàtics, U. Jaume I, 12071 Castelló, Spain |
| (4) |
Av. Tecnológico s/n, Instituto Tecnológico de Toluca, 52140 Metepec, México |
| (5) |
Dept. d’Informàtica, U. València, 46100 Burjassot (Valéncia), Spain |
Abstract
Filtering (or editing) is mainly effective in improving the classification accuracy of the Nearest Neighbour (NN) rule, and
also in reducing its storage and computational requirements. This work reviews some well-known editing algorithms for NN classification
and presents alternative approaches based on combining the NN and the Nearest Centroid Neighbourhood of a sample. Finally,
an empirical analysis over real data sets is provided.
This work has partially been supported by grants TIC2000-1703-C03-03 from the Spanish CICYT, SAB2001-0004 from the Spanish
MECD, and 32016-A from the Mexican CONACyT.
Fulltext Preview (Small, Large)
 References secured to subscribers.
|
|
|
|
|
|