Skip to main content
. 2018 May 9;12:272. doi: 10.3389/fnins.2018.00272

Table 9.

The mean classification accuracy and standard deviation (%) of 5 cross-validation (20% samples for training) with the number of corresponding channels (listed in the bracket) of COL and baselines on 5 subjects of DS1.

Method Subject Mean
aa al av aw ay
BS1 66.43 74.73 55.80 70.63 66.43 66.80
± 3.90 ± 1.21 ± 2.42 ± 2.63 ± 2.31
BS2 61.96 75.18 54.82 70.89 74.29 67.43
± 3.75 ± 7.10 ± 5.45 ± 3.32 ± 3.48
gsBLDA 64.64
(106)
80.18
(81)
56.70
(96)
73.75
(95)
68.84
(86)
68.82
± 5.26 ± 2.76 ± 5.28 ± 4.65 ± 5.01
MRCS 66.87
(115)
76.16
(71)
57.68
(49)
73.12
(75)
69.46
(33)
68.66
± 5.61 ± 1.08 ± 2.15 ± 3.21 ± 1.94
CSTI 64.73
(1)
89.64
(7)
59.82
(89)
68.30
(104)
85.80
(9)
73.66
± 13.52 ± 5.92 ± 3.09 ± 2.53 ± 5.50
NSGA-II 69.37
(47)
77.86
(39)
58.39
(45)
72.50
(60)
71.25
(34)
69.87
± 4.20 ± 2.29 ± 2.31 ± 3.44 ± 1.47
Proposed 80.43
(19)
90.88
(14.8)
68.43
(17.2)
94.13
(15.6)
93.18
(14.2)
84.81
± 3.34 ± 3.81 ± 5.29 ± 2.15 ± 2.00

The best performance is highlighted in bold. Performance of BS1 and BS2 indicates classification accuracy with all channels and 3C channels, respectively.