Feature selection and semi-supervised clustering using multiobjective optimization

SpringerPlus

Table 2 Results on different data sets by different algorithms

Data set	Semi-FeaClusMOO			Semi-FeaClusMOO_Euc			FeaClusMOO			VAMOSA		VGAPS		KM
	Fea	OC	MS	Fea	OC	MS	Fea	OC	MS	OC	MS	OC	MS	MS
Iris	4	3	0.39	2,3,4	4	0.40	3,4	3	0.44	2	0.80	3	0.62	0.68
Cancer	1,2,3,5,6	2	0.31	1,2,5,6,7	2	0.37	1,2,3,5,6	2	0.31	2	0.32	2	0.37	0.37
Newthy.	2,4	3	0.46	2-4	3	0.47	1,2,4,5	3	0.54	5	0.57	3	0.58	0.94
Wine	1,3,6,8,9,12	3	0.62	1,6,7,10,12	6	0.64	1,6,7	3	0.67	3	0.97	3	1.12	1.40
LiverDis.	1,6,7,10,12	2	0.64	1,2,3,6	3	0.98	1,2,5	2	0.98	2	0.98	2	0.98	0.98
LungCan.	1-4,	2	0.70	2-4	4	0.71	1-4,	2	0.70	3	0.85	3	1.24	1.45
	7-8,10,			6-9			7-8,10,
	11,13,16			11-14			11,13,16
	19-23,			16,22			19-23,
	25-27			27-32			25-27
	29-31			34,36-38			29-31
	32-39			41,43-46			32-39
	42-45			48,49			42-45
	47-49			51-56			47-49
	51-53						51-53
	55-56						55-56
Glass	1-4,7-9	6	1.03	4,9	10	1.05	1,2,3,4,5	6	1.05	6	1.08	6	1.10	1.69

Here d, K, Fea, OC, MS denote respectively the original dimension, the number of features selected by the algorithm, the number of clusters originally present, obtained number of clusters and Minkowski Score, respectively. Each algorithm is executed ten times and the best results among these ten runs are reported.