Table 3. Classification accuracy of the proposed method based on 30 different runs.

Number of experiments	Training accuracy (%)	Validation accuracy (%)	Testing accuracy (%)	Percent error
1	97.7	96.5	98.7	1.3
2	98.5	98.7	98.8	1.2
3	98.9	97.9	99.2	0.8
4	98.2	98.7	98.5	1.5
5	97.7	97.2	98.6	1.4
6	98.6	98.6	98.8	1.2
7	98.8	98.5	99.2	0.8
8	98.4	98.2	98.4	1.6
9	98.7	98.7	99.2	0.8
10	98.5	98.5	98.8	1.2
11	98.2	98.2	98.8	1.2
12	98.8	98.8	99.2	0.8
13	98.3	98.0	98.6	1.4
14	98.5	98.3	98.8	1.2
15	98.3	98.5	98.7	1.3
16	98.5	98.5	98.9	1.1
17	98.6	98.5	99.2	0.8
18	98.8	98.5	99.2	0.8
19	98.9	98.7	98.8	1.2
20	98.8	98.5	98.9	1.1
21	98.2	98.2	98.5	1.5
22	98.5	98.5	98.7	1.3
23	98.9	98.8	98.8	1.2
24	98.6	98.5	98.3	1.7
25	98.5	98.5	98.8	1.2
26	98.4	98.4	99.0	1.0
27	98.9	98.5	99.1	0.9
28	98.6	98.1	98.4	1.6
29	98.2	98.3	98.9	1.1
30	98.6	97.5	98.6	1.4
Average	98.5	98.3	98.8	1.2