A New Hierarchical Clustering Algorithm and Its Performance Validation Using High-Dimensional Artificial Data

Riichiro MIZOGUCHI  Hirotake NAKASHIMA  Osamu KAKUSHO  

Publication
IEICE TRANSACTIONS (1976-1990)   Vol.E66   No.6   pp.359-365
Publication Date: 1983/06/25
Online ISSN: 
DOI: 
Print ISSN: 0000-0000
Type of Manuscript: PAPER
Category: Pattern Recoguition
Keyword: 


Full Text: PDF(492.3KB)>>
Buy this Article




Summary: 
The present paper has the following three major objectives:
a) To propose a new hierarchical clustering algorithm based on a powerful distance measure defined by using k-nearest neighbor.
b) To discuss the validity problem of clustering.
c) To describe the need of the constructed data sets for validation of clustering algorithms and to execute the validation of several algorithms including ours using the data sets.
Section 2 describes the authors' view of cluster analysis and cluster itself where discussion is made lying stress on the necessity of establishing the concept of a cluster and developing an efficient clustering algorithm free from any restriction associated with the most existing ones. In Sect. 3, a new hierarchical clustering algorithm is proposed which can be considered as a modified version of single-link algorithm. The algorithm is based on a powerful distance measure defined by using the idea of k-nearest neighbor. In Sect. 4, some four-dimensional data for testing the validity of clustering algorithms is presented along with how to construct them. Section 5 describes some results of computer simulation of several clustering algorithms including ours using the constructed data. The results show the efficiency of out algorithm.