https://github.com/berndbischl/mlr
Tip revision: ce82484948ede7069cb0876916e6f993db2decf9 authored by pat-s on 26 November 2019, 22:20:58 UTC
use pkgdown dev & release
use pkgdown dev & release
Tip revision: ce82484
RLearner_cluster_Cobweb.R
#' @export
makeRLearner.cluster.Cobweb = function() {
makeRLearnerCluster(
cl = "cluster.Cobweb",
package = "RWeka",
par.set = makeParamSet(
makeNumericLearnerParam(id = "A", default = 1, lower = 0),
makeNumericLearnerParam(id = "C", default = 0.002, lower = 0),
makeIntegerLearnerParam(id = "S", default = 42L, lower = 1L)
),
properties = "numerics",
name = "Cobweb Clustering Algorithm",
short.name = "cobweb",
callees = c("Cobweb", "Weka_control")
)
}
#' @export
trainLearner.cluster.Cobweb = function(.learner, .task, .subset, .weights = NULL, ...) {
ctrl = RWeka::Weka_control(...)
RWeka::Cobweb(getTaskData(.task, .subset), control = ctrl)
}
#' @export
predictLearner.cluster.Cobweb = function(.learner, .model, .newdata, ...) {
# RWeka returns cluster indices (i.e. starting from 0, which some tools don't like
as.integer(predict(.model$learner.model, .newdata, ...)) + 1L
}