Table 5 Search space of the hyper-parameters explored for each model.

Random Survival Forest

number of trees

1 to 1000, logscale=TRUE^*

node depth

1 to 100

number of variables considered at each split (mtry)

1 to 36

split rule

{logrank, logrankscore}

Fast Random Survival Forest

min node size

1 to 100

number of variables considered at each split (mtry)

1 to 36

number of trees

1 to 1000, logscale=TRUE^*

Cvglmnet

alpha

(0,1)

lambda.min.ratio

(0,1)

Rpart

complexity parameter

(0,1)

DeepSurv

dropout

(0,1)

learning rate

(0,1)

DeepHit

dropout

(0,1)

learning rate

(0,1)

CoxTime

dropout

(0,1)

learning rate

(0,1)