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
|
test_that("classif_multinom", {
requirePackagesOrSkip("nnet", default.method = "load")
capture.output({
m = nnet::multinom(formula = multiclass.formula, data = multiclass.train)
})
p = predict(m, newdata = multiclass.test)
testSimple("classif.multinom", multiclass.df, multiclass.target, multiclass.train.inds, p)
p = predict(m, newdata = multiclass.test, type = "probs")
testProb("classif.multinom", multiclass.df, multiclass.target, multiclass.train.inds, p)
tt = nnet::multinom
tp = function(model, newdata) predict(model, newdata)
testCV("classif.multinom", multiclass.df, multiclass.target, tune.train = tt, tune.predict = tp)
# test multinom for 2 classes
wl = makeLearner("classif.multinom", predict.type = "prob")
m = train(wl, binaryclass.task)
p = predict(m, newdata = binaryclass.df)
rr = p$data$response
pp = getPredictionProbabilities(p)
i = as.integer(pp < 0.5) + 1
labs = as.factor(getTaskClassLevels(binaryclass.task)[i])
expect_equal(rr, labs)
})
|
