Skip to main content

Table 2 Binary classifier evaluation metrics on test set

From: Automated vetting of radiology referrals: exploring natural language processing and traditional machine learning approaches

Model

Weighted accuracy (%)

Sensitivity (%)

Specificity (%)

AUC

BOW + DSW + LR

88.3

86.7

90.0

0.925

BOW + DSW + SVM

92.8

88.9

96.7

0.942

BOW + DSW + RF

88.3

86.7

86.7

0.930

TF-IDF + DSW + LR

87.2

84.4

90.0

0.923

TF-IDF + DSW + SVM

86.1

88.9

83.3

0.923

TF-IDF + DSW + RF

85.0

86.7

86.7

0.931

BOW + CSW + LR

87.2

84.4

90.0

0.915

BOW + CSW + SVM

88.9

84.4

93.3

0.932

BOW + CSW + RF

85.6

84.4

86.7

0.910

TF-IDF + CSW + LR

85.0

80.0

90.0

0.917

TF-IDF + CSW + SVM

85.0

80.0

90.0

0.926

TF-IDF + CSW + RF

85.6

84.4

86.7

0.910

BOW + CSW + SC + LR

87.2

84.4

90.0

0.932

BOW + CSW + SC + SVM

92.2

91.1

93.3

0.948

BOW + CSW + SC + RF

87.2

84.4

90.0

0.911

  1. BOW bag-of-words, DSW default stop words, CSW custom stop words, LR logistic regression, RF random forest, SC spell checker, SVM support vector machine