Table 3 In simulation scenario 2–1, the Feature relative importance obtained through absolute sum normalization.
From: Pseudo datasets estimate feature attribution in artificial neural networks
Feature | LR coefficients | PDPE | SHAP value |
|---|---|---|---|
Mean (95% CI) | Mean (95% CI) | Mean (95% CI) | |
X1 | 0.2609 (0.2579, 0.2640) | 0.2483 (0.2448, 0.2515) | 0.5058 (0.4979, 0.5137) |
X2 | − 0.1555 (− 0.1583, − 0.1526) | − 0.1519 (− 0.1549, − 0.1488) | 0.1960 (0.1894, 0.2024) |
X3 | 0.1041 (0.1018, 0.1067) | 0.1045 (0.1018, 0.1068) | 0.0950 (0.0908, 0.0994) |
X4 | 0.0004 (− 0.0021, 0.0032) | 0.0002 (− 0.0028, 0.0029) | 0.0083 (0.0069, 0.0097) |
X5 | − 0.0002 (− 0.0026, 0.0021) | 0.0008 (− 0.0015, 0.0032) | 0.0075 (0.0061, 0.0089) |
X6 | 0.1826 (0.1777, 0.1874) | 0.1864 (0.1808, 0.1914) | 0.0744 (0.0698, 0.0790) |
X7 | 0.0262 (0.0214, 0.0304) | 0.0266 (0.0214, 0.0317) | 0.0073 (0.0062, 0.0085) |
X8 | − 0.2061 (− 0.2100, − 0.2022) | − 0.2116 (− 0.2161, − 0.2069) | 0.0933 (0.0883, 0.0981) |
X9 | − 0.0003 (− 0.0055, 0.0046) | 0.0000 (− 0.0050, 0.0050) | 0.0064 (0.0055, 0.0074) |
X10 | 0.0021 (− 0.0032, 0.0072) | 0.0028 (− 0.0027, 0.0082) | 0.0059 (0.0048, 0.0071) |
