1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
|
test_that("fgam works for classifcation", {
# errors on R 4.1
skip_if(sessionInfo()$R.version$status == "Under development (unstable)")
requirePackagesOrSkip("refund")
dd = getTaskData(gunpoint.task, functionals.as = "matrix", target.extra = TRUE)
matdd = list()
matdd$fd = dd$data$fd
hh = getBinomialTarget(gunpoint.task)
matdd$X1 = hh$newtarget
fit.af = pfr(formula = X1 ~ af(fd, Qtransform = TRUE, k = 3, m = 2), data = matdd, family = binomial())
lrn = makeLearner("classif.fgam", par.vals = list(mgcv.te_ti.k = 3L, mgcv.te_ti.m = 2))
m = train(lrn, gunpoint.task)
cp = predict(m, task = gunpoint.task)
expect_class(cp, "Prediction")
# prob output
lrn = makeLearner("classif.fgam", par.vals = list(mgcv.te_ti.k = 3L, mgcv.te_ti.m = 2), predict.type = "prob")
m2 = train(lrn, gunpoint.task)
cp2 = predict(m2, task = gunpoint.task)
expect_class(cp2, "Prediction")
})
|
