cvLoss {gbev} | R Documentation |
Performs k-fold cross-validation of number of boosting iterations.
cvLoss(object,k,random=F,loss="logLike")
object |
A fitted model of type gbev.object . |
k |
Is the k in k -fold cross-validation. |
random |
If TRUE then cross-validation is done by randomly sampling (without replacement)
the validation group, else the validation groups are determined from order
appearing in data. |
loss |
Can be logLike for binary regression (negative log-likelihood loss), or L2 for squared error loss. |
Returns a list containing,
iters
which are iterations the cross-validation loss is evaluated,
cvLoss
which is the cross-validated loss-function,
estIter
which is the iteration minimizing the cross-validated loss.