Skip to main content

Table 4 Variance of feature importances. Variance of the five most important features of a 10-fold cross-validation

From: Compensation of feature selection biases accompanied with improved predictive performance for binary classification by using a novel ensemble feature selection approach

Dataset Variance #1 Variance #2 Variance #3 Variance #4 Variance #5
MI-Mortality 0.001759124 0.004694053 0.004904828 0.003720571 0.001580310
Fibrosis 0.003124527 0.008085472 0.019901386 0.009202372 0.019804508
FLIP 0.006604973 0.011325453 0.014731007 0.023499884 0.020140657
SPECTF 0.000380482 0.014946809 0.011520607 0.005807655 0.002880478
Sonar 0.003887830 0.001792209 0.003004598 0.003115140 0.002680274
WBC 0.001071784 0.001769331 0.002912278 0.000387555 0.001096465