Journal of Big Data

Table 1 Characteristics of the datasets (d: dimensions, N: number of objects, K: number of clusters or classes)

From: An ensemble method for estimating the number of clusters in a big data set using multiple random samples

Dataset	d	Cluster/class sizes	N	K
DS1	2	3 clusters; each has 50,000	1,000,000	10
		4 clusters; each has 100,000
		3 clusters; each has 150,000
DS2	2	7 clusters; each has 25,000	1,000,000	20
		1 cluster; has 30,000
		1 cluster; has 45,000
		6 clusters; each has 50,000
		3 clusters; each has 75,000
		1 cluster; has 100,000
		1 cluster; has 125,000
DS3	10	3 clusters; each has 15,000	1,000,000	30
		3 clusters; each has 20,000
		11 clusters; each has 25,000
		1 cluster; has 30,000
		2 clusters; each has 35,000
		4 clusters; each has 40,000
		3 clusters; each has 50,000
		1 cluster; has 60,000
		2 clusters; each has 75,000
DS4	10	5 clusters; each has 5000	1,000,000	40
		2 clusters; each has 10,000
		4 clusters; each has 15,000
		4 clusters; each has 20,000
		16 clusters; each has 25,000
		1 cluster; has 35,000
		1 cluster; has 40,000
		5 clusters; each has 50,000
		1 cluster; has 60,000
		1 cluster; has 75,000
DS5	10	1 cluster; has 10,000	1,000,000	50
		19 clusters; each has 15,000
		14 clusters; each has 20,000
		13 clusters; each has 25,000
		1 cluster; has 30,000
		2 clusters; each has 35,000
Covertype	54	211,840 : 283,301 : 35,754 :	581,012	7
		27,747 : 8483 : 17,367 : 20,510
KDD’99ID	41	972,781 : 2,807,886 : 1,072,017 :	4,940,000	23
		53 : 979 : 264 : 21 : 2203 : 8 : 12 :
		2 : 1020 : 20 : 7 : 4 : 30 : 9 : 10 :
		3 : 12,481 : 15,892 : 2316 : 10,413
PokerHand	10	513,702 : 433,097 : 48,828: 3978 :	1,025,010	10
		21,634 : 2050 : 1460 : 236 : 17 : 8
SUSY	18	2,712,173 : 2,287,827	5,000,000	2

In the four real-world datasets, the K values in the middle column are the numbers of objects in K classes
The small classes are underlined

Back to article page