Channel and Feature Selection for a Motor Imagery-Based BCI System Using Multilevel Particle Swarm Optimization

Table 2

Parameter setting and classification result of two-level PSO for feature selection.

	PSO parameters							Results		Runtimes
No.	N	T						NF	ACC (%)	Training time (s)	Test time (s)
1	100	100	1.5	1.5	0.8	0.4	20	541	98	13,717	1.089
2	100	100	1.3	1.3	0.8	0.4	20	556	94	13,637	1.125
3	100	100	1.7	1.7	0.8	0.4	20	542	94	13,720	1.154
4	100	100	1.5	1.5	1.0	0.4	20	555	94	13,753	1.108
5	100	100	1.5	1.5	1.0	0.2	20	580	94	13,701	1.176
6	100	100	1.5	1.5	0.8	0.2	20	544	95	13,938	1.146
7	100	100	1.5	1.5	0.8	0.4	18	572	95	13,700	1.125
8	100	100	1.5	1.5	0.8	0.4	22	564	94	13,579	1.128
9	80	100	1.5	1.5	0.8	0.4	20	580	96	10,812	1.139
10	120	100	1.5	1.5	0.8	0.4	20	587	96	15,862	1.167

No.: number; N: population size; T: maximum number of iterations; NF: number of selected features; ACC: accuracy.