Evaluate by cross-validation (leave-one-out) the effect induced by leaving out each performance on the result of a functional clustering.
ftest_performances(fres, opt.nbMax = fres$nbOpt,
opt.R2 = FALSE, opt.plot = FALSE)an object resulting from a functional clustering
obtained with the whole dataset using the function fclust.
a logical. If opt.plot = TRUE,
at each test, the tree resulting from removing
each component, assemblage or performance is plotted.
a logical. If opt.R2 = TRUE,
the primary tree is validated
and the vectors of coefficient of determination (R^2)
and efficiency (E) are computed.
a logical. If opt.plot = TRUE,
at each test, the tree resulting from removing
each component, assemblage or performance is plotted.
a list containing a matrix by clustering index.
Each performance of the dataset is successively removed,
the remaining performance collection is functionally analysed,
then indices of distance between clustering trees
with and without the performance are computed.
The performances can then be hierarchised depending on
the distance induced by their removing from dataset.
The used distance criteria are :
"Czekanowski_Dice", "Folkes_Mallows", "Jaccard", "Kulczynski",
"Precision", "Rand", "Recall", "Rogers_Tanimoto", "Russel_Rao",
"Sokal_Sneath1" and "Sokal_Sneath2" index.
For more informations, see the notice of R-package clusterCrit.
The test is time-consuming.
Package "clusterCrit": Clustering Indices, by Bernard Desgraupes (University of Paris Ouest - Lab Modal'X)