validatetree {ConsRankClass}

Validation of the tree for preference rankings

Description

Validation of the tree either with a test set procedure or with v-fold cross validation

Usage

validatetree(
  Tree,
  testX = NULL,
  testY = NULL,
  method = "test",
  V = 5,
  plotting = TRUE
)

Arguments

Tree

An object of the class "ranktree" coming form te function ranktree

testX

The data frame containing the test set (predictors)

testY

The matrix ontaining the test set (response)

method

One between "test" (default) or "cv"

V

The cross-validation parameter. Default V=5

plotting

With the defaul option plotting=TRUE, the pruning sequence plot is visualized

Value

A list containing:

tau		the Tau_x rank correlation coefficient of the sequence of the trees
error		the error of the sequence of the trees
termnodes		the number of terminal nodes of the sequence of the trees
best_tau		the best tree in terms of Tau_x rank correlation coefficient
best_error		the best tree in terms of error (it is the same)
validation		information about the validation procedure

#'

Author(s)

Antonio D'Ambrosio [e-mail]

Examples

data(EVS)
EVS$rankings[is.na(EVS$rankings)] <- 3
set.seed(654)
training=sample(1911,1434)
tree <- ranktree(EVS$rankings[training,],EVS$predictors[training,],decrmin=0.001,num=50)
#test set validation
vtreetest <- validatetree(tree,testX=EVS$predictors[-training,],EVS$rankings[-training,]) 
#cross-validation
vtreecv <- validatetree(tree,method="cv",V=10)