Improved Instance Selection Methods for Support Vector Machine Speed Optimization
Table 4
Result for FFA_IS.
Total email
NFF
NI
APA (%)
GB
FP (%)
FN (%)
(%)
Pr (%)
FM (%)
(s)
4000
5
100
99.25
100
0.29
4.00
96.00
97.98
96.83
23.54
4000
5
150
99.55
100
0.20
2.20
97.80
98.67
98.11
39.07
4000
5
200
99.55
100
0.11
2.80
97.20
99.20
98.08
62.24
4000
5
250
99.60
100
0.17
2.00
98.00
98.83
98.32
79.12
4000
5
300
99.65
100
0.11
2.00
98.00
99.22
98.49
109.70
4000
10
100
99.38
100
0.17
3.80
96.20
98.87
97.28
28.57
4000
10
150
99.55
100
0.26
1.80
98.20
98.29
98.13
45.94
4000
10
200
99.63
100
0.23
1.40
98.60
98.46
98.46
84.30
4000
10
250
99.68
100
0.09
2.00
98.00
99.42
98.59
117.98
4000
10
300
99.60
100
0.17
2.00
98.00
98.83
98.30
213.17
Key: NFF: number of fireflies, NI: number of instances, GB: global best, APA: average prediction accuracy, FP: false positive, FN: false negative, : recall, Pr: precision, FM: -measure, and : time.