x=as.h2o(x)

lassification variable was a factor vector. H2O ran an inbuilt

n process. The following code was used for this inbuilt validation,

pha and beta were both fraction values between zero and one.

m should be smaller than one.

splits=h2o.splitFrame(x,c(alpha,beta))

raining function is shown below, where h1 and h2 were the

of hidden neurons in two layers, train.id and valid.id

ained by h2o.splitFrame, response was the classification

name in the data matrix, predictors was for all the

ent variables and epochs was the learning cycle,

eplearning(model_id='dl_model_first',

aining_frame=train.id,

lidation_frame=valid.id,

predictors,y=response,hidden=c(h1,h2),

pochs)

e 3.36(a) shows the ROC curve of a Dayhoff bio-DNN model

ed for the factor Xa protease cleavage data.

(a) (b)

a) The ROC curve of the bio-DNN model constructed for the factor Xa protease

ata. The AUC was 0.929. (b) The ROC curve of a mixture bio-DNN model

for the factor Xa data. The AUC was 0.958, which was the highest.