Using Hierarchical Clustering to Enhance Classification Accuracy
BibTeX
@MISC{Theodorakis_usinghierarchical,
author = {Matthaios Theodorakis and Andreas Vlachos and Theodore Z. Kalamboukis},
title = {Using Hierarchical Clustering to Enhance Classification Accuracy},
year = {}
}
Bookmark
 |
 |
 |
 |
 |
 |
OpenURL
Abstract
Abstract. A new approach to classification is presented based on COBWEB, an unsupervised conceptual clustering algorithm. The modifications proposed improved the classification accuracy by 2.32 % and up to 7.25 % in the Period Disambiguation system that was built in order to test the efficiency of the approach. The system can be trained across different domains and languages. It has been tested on the Brown Corpus and on a collection of articles from Greek financial newspapers achieving accuracy 99.18 % and 99.35 % respectively. 1
