Table 1 In simulation scenario 1–1, the values of feature importance obtained from XAI methods.

From: Pseudo datasets estimate feature attribution in artificial neural networks

Feature

LR coefficients

PDPE

SHAP Value

X1

0.0071 ± 0.0697

0.0045 ± 0.0250

0.0123 ± 0.0080

X2

0.0114 ± 0.0610

0.0082 ± 0.0241

0.0130 ± 0.0103

X3

0.0039 ± 0.0644

0.0072 ± 0.0239

0.0146 ± 0.0098

X4

0.0028 ± 0.0646

0.0063 ± 0.0203

0.0111 ± 0.0074

X5

− 0.0026 ± 0.0651

0.0059 ± 0.0244

0.0137 ± 0.0105

X6

− 0.0046 ± 0.1304

− 0.0067 ± 0.0418

0.0082 ± 0.0059

X7

0.0090 ± 0.1123

0.0033 ± 0.0392

0.0082 ± 0.0058

X8

0.0022 ± 0.1276

0.0004 ± 0.0421

0.0098 ± 0.0073

X9

− 0.0098 ± 0.1339

− 0.0082 ± 0.0464

0.0100 ± 0.0088

X10

0.0004 ± 0.1220

− 0.0072 ± 0.0383

0.0084 ± 0.0067

  1. LR, Logistic regression; PDPE, Pseudo datasets effect; SHAP, Shapley additive explanations.
Back to article page