hyperSMURF (version 2.0)

hyperSMURF.test.thresh: Test of a thresholded hyperSMURF model

Description

The predictions of each random forest are discrete, i.e. 1 or 0: the probabilities are thresholded according to the cutoff value set in the training phase. The threshold is embedded in the HSmodel according to the cutoff parameter set in the training phase. The score computed by the hyperensemble is the average of the discrete predictions generated by each base random forest.

Usage

hyperSMURF.test.thresh(data, HSmodel)

Arguments

data

a data frame or matrix with the test data. Rows: examples; columns: features

HSmodel

a list including the trained random forest models. The models have been trained with hyperSMURF.train. The threshold is embedded in the model according to the cutoff value set in the training phase.

Value

a named vector with the computed probabilities for each example (HyeprSMURF thresholded score)

See Also

hyperSMURF.test, hyperSMURF.train

Examples

# NOT RUN {
train <- imbalanced.data.generator(n.pos=20, n.neg=500, 
         n.features=10, n.inf.features=2, sd=0.1, seed=1);
HSmodel <- hyperSMURF.train(train$data, train$label, n.part = 5, 
                  fp = 1, ratio = 2, k = 5, cutoff=c(0.3, 0.7));
test <- imbalanced.data.generator(n.pos=20, n.neg=500, 
         n.features=10, n.inf.features=2, sd=0.1, seed=2);
res <- hyperSMURF.test.thresh(test$data, HSmodel);
# }