Table 3 Hyperparameter tuning via randomized search with tenfold cross-validation.

From: Personalized prediction of delayed graft function for recipients of deceased donor kidney transplants with machine learning

Model

ROC-AUC ± SD

Number of searches done

Best parameters

BL

0.703 ± 0.011

NA

None

LR

0.728 ± 0.012

NA

None

RF

0.735 ± 0.009

30

mtry = 3

EN

0.728 ± 0.012

100

alpha = 0.883, lambda = 0.00142

XGB

0.742 ± 0.009

100

nrounds = 668, max_depth = 6, eta = 0.0347, gamma = 5.703, subsample = 0.569, colsample_bytree = 0.699, rate_drop = 0.350, skip_drop = 0.805, min_child_weight = 7

ANN

0.737 ± 0.007

100

size = 20, decay = 8.795, number of layer = 1, entropy = TRUE, abstol = 1.0e−4, reltol = 1.0e−8, maxit = 1.0e6

  1. ROC-AUC area under the receiver operating characteristic curve, SD standard deviation, BL baseline, LR logistic regression, EN elastic net, RF random forest, XGB extreme gradient boosting, ANN artificial neural network.
Back to article page