Skip to main content

Table 2 LIME results for four people from the first fold on columns, nine most important features from logistic regression (LR) and then random forest (RF) on rows

From: Comparison of feature importance measures as explanations for classification models

Correctly classified bening LR Correctly classified malignant LR
Radius2 0.23 Concavity1 0.34
Fractal dimension2 − 0.21 Area2 0.25
Concave points2 0.19 Radius2 0.23
Fractal dimension3 0.18 Concave points1 0.18
Compactness1 − 0.17 Concave points2 0.17
Texture3 − 0.15 Radius3 0.15
Radius3 − 0.15 Texture3 0.15
Symmetry2 − 0.14 Fractal dimension2 0.13
Compactness2 − 0.13 Area3 0.12
Misclassified bening LR Misclassified malignant LR
Concavity1 0.33 Texture3 0.16
Fractal dimension2 − 0.24 Symmetry3 − 0.11
Compactness1 − 0.18 Symmetry2 0.08
Concave points2 0.18 Fractal dimension3 0.07
Fractal dimension3 0.18 Symmetry1 0.06
Symmetry2 − 0.15 Concave points2 − 0.04
Symmetry3 0.14 Texture2 -0.04
Compactness2 − 0.14 Concave points3 − 0.03
Texture3 0.14 Compactness1 − 0.03
Correctly classified bening RF Correctly classified malignant RF
Area3 − 0.08 Perimeter3 0.14
Perimeter3 − 0.08 Area3 0.13
Radius3 − 0.07 Concave points3 0.13
Concave points3 − 0.07 Radius3 0.12
Texture3 − 0.05 Area2 0.07
Concave points1 − 0.03 Concave points1 0.06
Concavity3 − 0.03 Texture3 0.05
Area2 − 0.03 Area1 0.05
Texture1 − 0.02 Texture1 0.04
Misclassified bening RF Misclassified malignant RF
Perimeter3 − 0.08 Area3 0.14
Area3 − 0.08 Perimeter3 0.14
Radius3 − 0.06 Radius3 0.12
Texture3 0.06 Concave points3 − 0.07
Concavity3 0.04 area2 0.06
Area2 − 0.04 Area1 0.05
Smoothness3 0.02 Texture3 − 0.05
Area1 − 0.01 Texture1 − 0.04
Concave points1 0.01 Concavity3 − 0.03
  1. Bolded are those features that were detected also by both classification methods