Skip to main content

Table 1 Performance of SVM classifiers for various combinations of training features, kernels, parameters and validation methods

From: LipocalinPred: a SVM-based method for prediction of lipocalins

Feature

V*

Kernel

Parameters

SN (%)

SP (%)

Acc (%)

MCC

F measure

   

Threshold

C

γ

d

     

AAC

A

R

-0.1

1

0.01

-

72.79

77.10

75.16

0.498

1.429

DPC

A

P

-0.1

0

-

2

80.14

87.34

84.10

0.678

1.658

PSSM

A

R

-0.1

5

9

-

89.70

89.15

89.40

0.786

1.725

 

D

R

-0.1

5

9

-

84.55

85.54

85.09

0.701

1.644

 

D

R

-0.1

5

9

-

88.96

84.33

86.42

0.731

1.633

SSC

A

R

-0.1

5

3

-

86.02

86.74

86.42

0.726

1.665

 

D

R

-0.1

5

3

-

84.55

86.74

85.75

0.712

1.664

 

D

R

-0.1

5

3

-

82.35

78.91

80.46

0.609

1.509

DPC+SSC

A

P

0.1

0

-

2

85.29

86.14

85.76

0.713

1.651

PSSM+SSC

A

R

0.0

4

1

-

88.97

92.16

90.72

0.812

1.785

 

A

R

-0.1

4

1

-

89.70

89.15

89.40

0.786

1.725

 

D

R

0.0

4

1

-

87.49

80.72

83.77

0.678

1.561

 

D

R

0.0

4

1

-

85.29

84.93

85.09

0.700

1.628

DPC+PSSM

A

R

-0.1

0

0.001

-

81.61

83.73

82.78

0.652

1.592

DPC+PSSM+SSC

A

P

0.1

0

-

2

85.29

86.14

85.76

0.713

1.651

  1. *Validation: A = Leave-one-out; D = Hold-out
  2. Kernel: R = RBF; P = Polynomial; L = Linear