Visual comparison of clustering using link-based clustering method (Lbcm) without predetermining initial centroid information

Document Type

Article

Source of Publication

ICIC Express Letters, Part B: Applications

Publication Date

4-1-2021

Abstract

High dimensional data are difficult to view in two-dimensional plot. However, having a mechanism to reduce to a selected number of salient features that can well present the data is essential. We attempted to reduce N dimensional data to two-dimensional data using the combination of Information Gain (IG) and Principal Component Analysis (PCA) and to perform the link-based clustering which is our novel technique presented in this work in determining the linked clusters automatically using visual approach. Link-based Clustering Method (LbCM) is applied on the two-dimensional data to determine the clusters automatically. The significance of the method is that it does not require prior information such as the number of linked clusters. The approach using a combination of IG-PCA for feature selection is also useful to deal with high dimensional data. The LbCM is able to detect the number of linked clusters automatically by analyzing the X-Y coordinate positions of the points and visual information such as gaps between points and of two extreme points for both axes. Since the number of clusters is represented visually in two dimensions, LbCM performance can be compared visually.

ISSN

2185-2766

Volume

12

Issue

4

First Page

317

Last Page

323

Disciplines

Computer Sciences

Keywords

Clustering algorithm, Density-based clustering, Information gain, Principal component analysis

Scopus ID

85102534859

Indexed in Scopus

yes

Open Access

no

Link to Full Text

Share

COinS