Improved sqrt-cosine similarity measurement

Journal of Big Data

Table 11 Performance of the similarity measures in data sets averaged across all classification learners

Metric	Data set	ISC		Cosine		Gaussian
Metric	Data set	Mean	HSD	Mean	HSD	Mean	HSD
Accuracy	WEBKB	0.6104	A	0.5929	A	0.3046	A
	R8	0.7166	A	0.7361	A	0.4485	A
	R52	0.4975	A	0.4230	A	0.1945	A
	NEWS	0.6009	A	0.5989	A	0.2468	A
	DBLP	0.7101	A	0.6842	A	0.2234	B
	CSTR	0.8019	A	0.7873	A	0.3052	B
Average #A’s		0.6562	6	0.6370	6	0.2871	4
AUC	WEBKB	0.8162	A	0.8304	A	0.6171	A
	R8	0.7342	A	0.6641	A	0.5341	A
	R52	0.7826	A	0.7540	A	0.5075	A
	NEWS	0.7514	A	0.7570	A	0.5852	A
	DBLP	0.9253	A	0.9287	A	0.6011	B
	CSTR	0.7313	A	0.7340	A	0.5040	A
Average #A’s		0.7901	6	0.7780	6	0.5581	5