Table 2. Clustering of 10 datasets (generated after performing 10 attribute weighting algorithms) into T (mesophile) and F (thermophile) classes by four different unsupervised clustering algorithms (K-Means, K-Medoids, SVC and EMC).

	Chi Squared		Correlation		Deviation		Gini Index		Information Gain		Relief		Rule		PCA		SVM		Uncertainty
	T	F	T	F	T	F	T	F	T	F	T	F	T	F	T	F	T	F	T	F
K-Means	1461	596	1810	247	1222	835	1452	605	1333	724	1603	454	1601	456	434	1623	1076	981	372	1685
K-Medoids	487	1570	1521	536	104	1953	1570	487	1152	905	583	1474	1652	405	1768	289	892	1165	939	1118
SVC	363	1688	1701	328	1705	6	363	1688	570	1487	529	1324	1561	4	631	1426	0	2057	1089	947
EMC	0	2057	1544	513	0	0	0	2057	0	2057	0	2057	4	2053	0	2057	0	2057	1544	513

The actual numbers of T (mesostable) and F (thermostable) classes in the original datasets were 1544 and 513, respectively. The highest accuracy (100%) was observed when the EMC clustering method was applied to datasets generated by Correlation and Uncertainty attribute weighting algorithms that highlighted in the table.