Table 13. Performance evaluation of all measures when NF = the whole size (Reuters = 18,308, web-kb = 33,025 features) – the averaged results (K = 1–120; +2).

Dataset	Reuters						Web-KB
Similarity/criterion	ACC	PRE	REC	FM	GM	AMP	ACC	PRE	REC	FM	GM	AMP
Euclidean	0.882	0.895	0.779	0.821	0.870	0.706	0.760	0.773	0.712	0.714	0.808	0.604
Cosine	0.884	0.898	0.779	0.823	0.871	0.709	0.760	0.773	0.712	0.714	0.808	0.604
Ex-Jaccard	0.904	0.883	0.666	0.700	0.809	0.583	0.760	0.800	0.679	0.685	0.787	0.591
Manhattan	0.527	0.378	0.167	0.151	0.379	0.164	0.398	0.223	0.258	0.155	0.441	0.256
KLD	0.301	0.255	0.129	0.065	0.337	0.128	0.387	0.177	0.251	0.140	0.434	0.250
SMTP	0.905	0.894	0.688	0.733	0.821	0.617	0.804	0.851	0.718	0.730	0.816	0.644
BLAB-SM	0.902	0.848	0.659	0.708	0.803	0.590	0.814	0.865	0.723	0.731	0.820	0.656