Feature selection using angle modulated simulated Kalman filter for peak classification of EEG signals

SpringerPlus

Table 6 Best testing results over 30 runs using the proposed AMSKF feature selection algorithm on eye event-related EEG data

Run	Training (%)	Validation (%)	Testing (%)	Best peak model														Feature subset length
1	87.52	63.88	69.19	1	3	4	6	7	8	9	10	11	12	13	14	15	16	14
2	90.14	63.92	62.89	1	2	3	4	5	6	7	8	9	15	16				11
3	95.12	61.30	55.78	1	2	3	4	5	6	7	8	9	10	11	16			12
4	91.77	61.68	72.71	1	2	7	8	9	10	11	12	13	14	15				11
5	78.33	65.99	56.51	13	14	15	16											4
6	89.44	71.36	62.21	3	6	7												3
7	93.81	67.50	66.78	1	2	8	9	10	11	12	13	14	15	16				11
8	96.61	67.19	60.02	1	5	9	13											4
9	94.65	64.64	66.50	1	2	14	15											4
10	92.20	60.68	57.87	2	3	8	9	10	13	14								7
11	95.74	66.54	62.55	1	11	12	15											4
12	82.57	65.36	61.47	12	13	14	15	16										5
13	92.20	71.06	64.64	1	2	5	13	14	15	16								7
14	91.50	71.13	59.16	3	6	14												3
15	89.44	58.06	60.60	1	2	3	7	8	10	11	13	15	16					10
16	88.19	65.65	60.32	1	2	5	6	7	8	9	10	13	14	15	16			12
17	90.83	70.24	55.20	1	2													2
18	86.92	67.34	60.51	1	2	5	6	7	8	9	10	13	14	15				11
19	95.24	62.63	61.98	1	2	3	4											4
20	88.80	68.93	66.51	1	2	3	15	16										5
21	85.54	66.92	61.66	9	10	11	12	13	14	15	16							8
22	94.15	66.02	57.85	1	3	4	7	9	11	14	16							8
23	82.12	62.33	61.34	12	13	14	15	16										5
24	95.59	65.14	62.30	1	2	3	9	10										5
25	83.67	68.40	62.37	1	2													2
26	92.08	66.54	61.75	3	9	15	16											4
27	80.18	63.01	61.96	14	15	16												3
28	94.15	66.95	52.96	1	10	11	12	13	14									6
29	87.60	60.47	63.47	12	13	14	15	16										5
30	89.92	71.94	62.34	3	4													2

The best-generalized peak model based on the maximum accuracy of testing data over 30 runs was marked with the italic font