Initial Seed Selection for Mixed Data Using Modified K-means Clustering Algorithm

Data sets to which clustering is applied may be homogeneous (numerical or categorical) or heterogeneous (numerical and categorical) in nature. Handling homogeneous data is easier than heterogeneous data. We propose a novel technique for identifying initial seeds for heterogeneous data clustering, th...

Full description

Saved in:
Bibliographic Details
Published in:Arabian journal for science and engineering (2011) Vol. 45; no. 4; pp. 2685 - 2703
Main Authors: Sajidha, S. A., Desikan, Kalyani, Chodnekar, Siddha Prabhu
Format: Journal Article
Language:English
Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.04.2020
Springer Nature B.V
Subjects:
Algorithms
Cluster analysis
Clustering
Datasets
Distance measurement
Engineering
Humanities and Social Sciences
multidisciplinary
Research Article - Computer Engineering and Computer Science
Science
Statistical tests
Vector quantization
means
Mixed attributes
Initial seed points
prototypes
Clustering
ISSN:2193-567X, 1319-8025, 2191-4281
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Data sets to which clustering is applied may be homogeneous (numerical or categorical) or heterogeneous (numerical and categorical) in nature. Handling homogeneous data is easier than heterogeneous data. We propose a novel technique for identifying initial seeds for heterogeneous data clustering, through the introduction of a unique distance measure where the distance of the numerical attributes is scaled such that it is comparable to that of categorical attributes. The proposed initial seed selection algorithm ensures selection of initial seed points from different clusters of the clustering solution which are then given as input to the modified K -means clustering algorithm along with the data set. This technique is independent of any user-defined parameter and thus can be easily applied to clusterable data sets with mixed attributes. We have also modified the K -means clustering algorithm to handle mixed attributes by incorporating our novel distance measure to handle numerical data and assigned the value one or zero when categorical data is dissimilar or similar. Finally, a comparison has been made with existing algorithms to bring out the significance of our approach. We also perform a statistical test to evaluate the statistical significance of our proposed technique.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:2193-567X
1319-8025
2191-4281
DOI:10.1007/s13369-019-04121-0