Gene Selection via a New Hybrid Ant Colony Optimization Algorithm for Cancer Classification in High-Dimensional Data

Table 8

A comparison between our method (MWIS-ACO-LS) and methods of state-of-the-art.

Datasets	Method	MWIS-ACO-LS	RBPSO-1NN	FBPSO-SVM	FRBPSO	HICATS	EPSO	TS-BPSO	IBPSO	IBPSO
		This work	2018	2018	2017	2016	2013	2009	2008	2011
	Performances		[6]	[6]	[21]	[28]	[25]	[23]	[22]	[24]
11_Tumors	Best #Acc (%)	99.42	—	—	—	97.70	96.55	97.35	93.10	95.4
	Best #Genes	101	—	—	—	287	243	3206	2948	228
	Average #Acc (%)	99.14 <1>	—	—	—	95.86	95.40	—	—	95.06
	Average #Genes	166.9	—	—	—	307.5	237.70	—	—	240.9
9_Tumors	Best #Acc (%)	100.00	83.33	95.00	—	83.33	76.67	81.63	78.33	78.33
	Best #Genes	40	20	71	—	259	251	2941	1280	248
	Average #Acc (%)	100.00 <1>	81.83	92.222	—	78.33	75.00	—	—	75.5
	Average #Genes	51	29.1	45	—	248.5	247.10	—	—	240.6
Brain_Tumor1	Best #Acc (%)	100.00	94.44	97,78	—	94.44	93.33	95.89	94.44	93.33
	Best #Genes	19	11	21	—	6	8	2913	754	5
	Average #Acc (%)	99.22 <1>	94.00	97.22	90.67	93.10	92.11	—	—	92.56
	Average #Genes	22.9	24.7	22.4	803	8.9	7.5	—	—	11.2
Brain_Tumor2	Best #Acc (%)	100.00	96.00	100.00	—	94.00	94.00	92.65	94.00	94.00
	Best #Genes	11	15	12	—	3	4	5086	1197	4
	Average #Acc (%)	99.40 <2>	92.80	100.00	87.6	92.60	92.4	—	—	91.00
	Average #Genes	11.1	24.5	14.3	662	5.8	6.0	—	—	6.4
Leukemia1	Best #Acc (%)	100.00	100.00	100.00	—	100.00	100.00	100.00	100.00	100.00
	Best #Genes	5	8	6	—	3	2	2577	1034	2
	Average #Acc (%)	100.00 <1>	99.72	100.00	98.89	100.00	100.00	—	—	100.00
	Average #Genes	9.4	11.7	8.4	825	3	3.2	—	—	3.2
Leukemia2	Best #Acc (%)	100.00	100.00	100.00	—	100.00	100.00	100.00	100.00	100.00
	Best #Genes	11	5	6	—	5	4	5609	1292	4
	Average #Acc (%)	100.00 <1>	100.00	100.00	97.50	100.00	100.00	—	—	100.00
	Average #Genes	13.9	13.1	8.6	1028	6.80	6.8	—	—	6.7
Lung_Cancer	Best #Acc (%)	99.51	—	—	—	97.04	96.06	99.52	96.55	96.55
	Best #Genes	36	—	—	—	7	7	6958	1897	10
	Average #Acc (%)	98.92 <1>	—	—	—	96.16	95.67	—	—	95.86
	Average #Genes	34.8	—	—	—	7.8	8.3	—	—	14.9
SRBCT	Best #Acc (%)	100.00	100.00	100.00	—	100.00	100.00	100.00	100.00	100.00
	Best #Genes	6	7	10	—	9	7	1084	431	6
	Average #Acc (%)	100.00 <1>	100.00	100.00	98.19	100.00	99.64	—	—	100.00
	Average #Genes	7.6	11.7	12.4	213	11.7	14.90	—	—	17.5
Prostate_Tumor	Best #Acc (%)	100.00	99.02	100.00	—	98.04	99.02	95.45	92.61	98.04
	Best #Genes	21	9	6	—	5	5	5320	1294	7
	Average #Acc (%)	99.12 <2>	98.24	100.00	92.43	97.75	97.84	—	—	97.94
	Average #Genes	20.3	11.2	8.3	418	7.2	6.6	—	—	13.6
DLBCL	Best #Acc (%)	100.00	100.00	100.00	—	100.00	100.00	100.00	100.00	100.00
	Best #Genes	6	6	4	—	3	3	2671	1042	4
	Average #Acc (%)	100.00 <1>	100.00	100.00	96.49	100.00	100.00	—	—	100.00
	Average #Genes	7.2	12.5	6.7	105	4.10	4.70	—	—	6

< >: the rank of our method in a specific average accuracy. RBPSO-1NN = a gene selection method based on the combination of ReliefF and BPSO and 1NN as a classifier. FBPSO-SVM = a gene selection method based on the combination of Fisher score and BPSO and the SVM as a classifier; FRBPSO = a fuzzy rule based binary PSO; HICATS=Hybrid Binary Imperialist Competition Algorithm and Tabu Search; EPSO = an enhancement of binary particle swarm optimization; TS-BPSO = A combination of tabu search and BPSO; IBPSO = an improved binary PSO.