1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
|
#' @export
makeRLearner.classif.logreg = function() {
makeRLearnerClassif(
cl = "classif.logreg",
package = "stats",
par.set = makeParamSet(
makeLogicalLearnerParam("model", default = TRUE, tunable = FALSE)
),
par.vals = list(
model = FALSE
),
properties = c("twoclass", "numerics", "factors", "prob", "weights"),
name = "Logistic Regression",
short.name = "logreg",
note = "Delegates to `glm` with `family = binomial(link = 'logit')`. We set 'model' to FALSE by default to save memory.",
callees = "glm"
)
}
#' @export
trainLearner.classif.logreg = function(.learner, .task, .subset, .weights = NULL, ...) {
f = getTaskFormula(.task)
# logreg expects the first label to be the negative class, contrary
# to the mlr3 convention that the positive class comes first.
# see https://github.com/mlr-org/mlr/issues/2817
tn = getTaskTargetNames(.task)
data = getTaskData(.task, .subset)
data[[tn]] = swap_levels(data[[tn]])
stats::glm(f, data = getTaskData(.task, .subset), family = "binomial", weights = .weights, ...)
}
#' @export
predictLearner.classif.logreg = function(.learner, .model, .newdata, ...) {
x = predict(.model$learner.model, newdata = .newdata, type = "response", ...)
levs = .model$task.desc$class.levels
if (.learner$predict.type == "prob") {
propVectorToMatrix(x, levs)
} else {
levs = .model$task.desc$class.levels
p = as.factor(ifelse(x > 0.5, levs[2L], levs[1L]))
unname(p)
}
}
|
