Skip to main content

Table 3 Prediction Performance (%): Feature Selection Methods Applied to the Whole Dataset (The last column lists the mean and standard deviation of the prediction performance over the permutation test (T = 100))

From: Ovarian cancer detection from metabolomic liquid chromatography/mass spectrometry data by support vector machines

Classifier

Feature Selection

52-20-split Validation

(50 trials)

12-fold CV

(10 trials)

LOOCV

    

Accuracy

Permutation Test Results

SVM

SVMRFE

91.1

94.2

95.8

97.7 ± 1.8

SVM

L1SVM

84.8

92.1

93.1

81.5 ± 7.1

SVM_NL

SVMRFE_NL

88.7

94.3

97.2

92.4 ± 3.5

SVM_NL

SVMRW

79.4

86.8

91.7

78.3 ± 6.8