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

Table 4

Classification results with different levels of PSO for feature selection.

Level	SF	ACC (%)	TRT (s)	TET (s)	Kappa	F-score	SEN	SPE	PRE
0^a	2,240	89	46	2.268	0.78	89.72	84.21	95.35	96
1	1,098	93	8,743	1.455	0.86	93.33	89.09	97.78	98
2	541	98	13,717	1.089	0.96	98	98	98	98
3	276	98	13,835	0.255	0.96	98	98	98	98
4	133	98	14,227	0.163	0.96	98	98	98	98
5	69	98	15,263	0.122	0.96	98	98	98	98
6	40	99	16,373	0.104	0.98	98.99	98	100	100
7	23	99	17,840	0.105	0.98	98.99	98	100	100
8	20	99	17,917	0.099	0.98	98.99	98	100	100
9	16	99	18,075	0.096	0.98	98.99	98	100	100
10	16	99	18,116	0.096	0.98	98.99	98	100	100

^aWithout PSO. SF: selected features; ACC: accuracy; TRT: training time; TET: test time; SEN: sensitivity; SPE: specificity; PRE: precision.