A Novel Hierarchical Clustering Approach Based on Universal Gravitation

Table 2

Well-tuned parameter configurations for different synthetic and real-world datasets.

Dataset	K-means	K-means++	SC		DBSCAN		Birth		LGC			GHC
	τ	τ	τ		m	e	t	b	i	n	f	γ	θ	k
Figure 4(a)	2	2	2	10	5	0.3	0.1	10	40	7	0.3	0.2	0.1	6
Figure 4(b)	7	7	7	20	20	0.2	0.5	10	30	10	0.3	0.2	0.1	6
Figure 4(c)	4	4	4	10	10	0.2	0.5	20	20	5	0.1	0.2	0.1	6
Figure 4(d)	5	5	5	70	4	0.1	0.4	40	20	10	0.1	0.2	0.1	6
Figure 4(e)	5	5	5	10	4	0.1	0.8	10	30	4	0.7	0.2	0.1	6
Figure 4(f)	3	3	3	20	10	0.2	0.1	10	20	10	0.1	0.2	0.1	6
Figure 4(g)	4	4	4	10	10	0.2	0.1	20	20	4	0.1	0.2	0.1	6
Figure 4(h)	2	2	2	10	10	0.3	0.1	100	20	4	0.1	0.2	0.1	6
Figure 4(i)	2	2	2	10	10	0.2	0.1	40	20	4	0.1	0.2	0.1	6
Figure 4(j)	2	2	2	80	5	0.2	0.4	10	5	4	0.3	0.2	0.1	6
BTissue	6	6	6	5	4	0.7	0.2	10	5	10	0.3	0.3	0.1	4
Ecoli	8	8	8	0.01	8	0.9	0.7	20	8	4	0.5	0.2	0.5	4
Iris	3	3	3	0.5	4	0.7	0.5	20	5	7	0.9	0.4	0	4
Wine	3	3	3	0.01	5	0.5	1	40	5	5	0.3	0.1	0.1	4
PID	2	2	2	0.5	10	0.2	0.2	20	20	4	0.9	0.1	0.1	6
BSTC	2	2	2	150	6	0.8	0.8	10	70	4	0.9	0.1	0.1	5
Glass	6	6	6	0.05	4	0.9	0.6	20	20	7	0.3	0	0.3	5
Cancer	2	2	2	0.01	4	0.9	0.7	50	30	4	0.7	0	0	6
SControl	6	6	6	150	10	10	0.5	30	5	4	0.9	0	0.3	5