Improved sqrt-cosine similarity measurement

Journal of Big Data

Table 12 Performance of the similarity measures in data sets averaged across all clustering learners

Metric	Data set	ISC		Cosine		Gaussian
Metric	Data set	Mean	HSD	Mean	HSD	Mean	HSD
Accuracy	WEBKB	0.4798	A	0.4434	A	0.3824	A
	R8	0.4384	A	0.4291	A	0.4472	A
	R52	0.2320	A	0.2283	A	0.2395	A
	NEWS	0.1659	A	0.1544	A	0.1179	A
	DBLP	0.3886	A	0.3574	A	0.2640	A
	CSTR	0.4332	A	0.4095	A	0.3182	A
Average #A’s		0.3563	6	0.3370	6	0.2948	6
Purity	WEBKB	0.6248	A	0.6091	A	0.5548	A
	R8	0.5769	A	0.5790	A	0.6446	A
	R52	0.4440	A	0.4225	A	0.4478	A
	NEWS	0.4234	A	0.4410	A	0.3948	A
	DBLP	0.7980	A	0.6363	A	0.6531	A
	CSTR	0.7026	A	0.6704	A	0.6700	A
Average #A’s		0.59495	6	0.55971	6	0.56085	6
NMI	WEBKB	0.1500	A	0.1177	A	0.0879	A
	R8	0.1978	A	0.1912	A	0.2030	A
	R52	0.1376	A	0.1321	A	0.1179	A
	NEWS	0.0855	A	0.0761	A	0.0731	A
	DBLP	0.2439	A	0.1940	A	0.0948	A
	CSTR	0.1396	A	0.1069	A	0.0172	A
Average #A’s		0.1590	6	0.1363	6	0.0990	6