testOutcomes
:"factor"
that
lists the actual outcomes in the records on the test set testPredictions
:"factor"
that
lists the predictions of outcomes in the test set testScores
:"ANY"
-- this
element will include matrices or vectors or arrays that include
information that is typically related to the posterior probability
of occupancy of the predicted class or of all classes. The actual
contents of this slot can be determined by inspecting the converter
element of the learnerSchema used to select the model. trainOutcomes
:"factor"
that
lists the actual outcomes in records on the training set trainPredictions
:"factor"
that
lists the predicted outcomes in the training set trainScores
:"ANY"
see
the description of testScores
above; the same information
is returned, but applicable to the training set records.
trainInd
:"numeric"
with of
indices of data to be used for training. RObject
:"ANY"
-- when
the trainInd
parameter of the MLearn
call is
numeric, this slot holds
the return value of the underlying R function that carried out
the predictive modeling. For example, if rpartI
was used
as MLearn method
, Robject
holds an instance of the
rpart
S3 class, and plot
and text
methods
can be applied to this. When the trainInd
parameter
of the MLearn
call is an instance of
xvalSpec
, this slot holds a list
of
results of cross-validatory iterations. Each element of this
list has two elements: test.idx
, giving the numeric
indices of the test cases for the associated cross-validation
iteration, and mlans
, which is the classifierOutput
for the associated iteration. See the example for an illustration
of 'digging out' the predicted probabilities associated with each
cross-validation iteration executed through an xvalSpec specification. embeddedCV
:fsHistory
:learnerSchema
:call
:"call"
-- records the
call used to generate the classifierOutput RObject signature(obj = "classifierOutput")
: Compute
the confusion matrix for test records. signature(obj = "classifierOutput")
: Compute
the confusion matrix for training set. Typically yields optimistically biased
information on misclassification rate. signature(obj = "classifierOutput")
: The R object
returned by the underlying classifier. This can then be passed on to
specific methods for those objects, when they exist. signature(obj = "classifierOutput")
: Returns
the indices of data used for training. signature(object = "classifierOutput")
: A print method
that provides a summary of the output of the classifier. signature(object = "classifierOutput")
: Print
the predicted classes for each sample/individual. The predictions
for the training set are the training outcomes. signature(object = "classifierOutput", t
= "numeric")
: Print the predicted classes for each
sample/individual that have a testScore
greater or equal
than t
. The predictions for the training set are the
training outcomes. Non-predicted cases and cases that
matche multiple classes are returned as NA
s.signature(object = "classifierOutput")
:
Returns the scores for predicted class for each
sample/individual. The scores for the training set are set to 1. signature(object = "classifierOutput")
:
Returns the prediction scores for all classes for each
sample/individual. The scores for the training set are set to 1 for
the appropriate class, 0 otherwise.signature(object = "classifierOutput")
: ... signature(object = "classifierOutput")
: Print
the predicted classes for each sample/individual in the test set. signature(object = "classifierOutput", t
= "numeric")
: Print the predicted classes for each
sample/individual in the test set that have a testScore
greater or equal than t
. Non-predicted cases and cases that
matche multiple classes are returned as NA
s.signature(object = "classifierOutput")
: ... signature(object =
"classifierOutput")
: Print the predicted classes for each
sample/individual in the train set. signature(object = "classifierOutput", t
= "numeric")
: Print the predicted classes for each
sample/individual in the train set that have a testScore
greater or equal than t
. Non-predicted cases and cases that
matche multiple classes are returned as NA
s.signature(object = "classifierOutput")
: ... showClass("classifierOutput")
library(golubEsets)
data(Golub_Train) # now cross-validate a neural net
set.seed(1234)
xv5 = xvalSpec("LOG", 5, balKfold.xvspec(5))
m2 = MLearn(ALL.AML~., Golub_Train[1000:1050,], nnetI, xv5,
size=5, decay=.01, maxit=1900 )
testScores(RObject(m2)[[1]]$mlans)
alls = lapply(RObject(m2), function(x) testScores(x$mlans))
Run the code above in your browser using DataLab