Clustering Analysis of Multivariate Data: A Weighted Spatial Ranks-Based Approach

Table 1

Comparison of different clustering approaches applied to the iris dataset.

Clustering method	No. of clusters	Cluster sizes	H	Purity	Entropy	ARI
WSR	2	50, 100	0	1	0	1
GMM (mclust) “BIC”	2	50, 100	0	1	0	1
K-means “CH index”	3	38, 62, 50	0.25	1	0	0.59
HDDC “BIC”	3	45, 55, 50	0.24	1	0	0.65
MixtPPCA “BIC”	3	50, 52, 48	0.32	1	0	0.57
PAM “silhouette width”	2	51, 99	0.01	0.99	0.05	0.97
DBSCAN	2	50, 97 (3 noise points)	0.02	1	0	0.5
KMD	3	50, 42, 58	0.28	1	0	0.58
FCM	3	60, 50, 40	0.27	1	0	0.58
GG	3	50, 65, 35	0.23	1	0	0.61
DDC	3	57, 57, 36	0.29	0.95	0.20	0.47
SNN	3	47, 94, 9	0.06	0.98	0.06	0.87
densityClust	2	50, 100	0	1	0	1

Results are based on the mean of 1000 repetitions.