+22
-17
lines changedFilter options
+22
-17
lines changed Original file line number Diff line number Diff line change
@@ -33,7 +33,7 @@ utils::globalVariables(
33 33
'lab', 'original', 'predicted_label', 'prediction', 'value', 'type',
34 34
"neighbors", ".submodels", "has_submodel", "max_neighbor", "max_penalty",
35 35
"max_terms", "max_tree", "name", "num_terms", "penalty", "trees",
36 -
"sub_neighbors")
36 +
"sub_neighbors", ".pred_class")
37 37
)
38 38 39 39
# nocov end
Original file line number Diff line number Diff line change
@@ -12,8 +12,10 @@
12 12
#' @param ... Optional arguments to pass to `predict.model_fit(type = "raw")`
13 13
#' such as `type`.
14 14
#' @return A tibble with the same number of rows as the data being predicted.
15 -
#' Mostly likely, there is a list-column named `.pred` that is a tibble with
16 -
#' multiple rows per sub-model.
15 +
#' There is a list-column named `.pred` that contains tibbles with
16 +
#' multiple rows per sub-model. Note that, within the tibbles, the column names
17 +
#' follow the usual standard based on prediction `type` (i.e. `.pred_class` for
18 +
#' `type = "class"` and so on).
17 19
#' @export
18 20
multi_predict <- function(object, ...) {
19 21
if (inherits(object$fit, "try-error")) {
@@ -404,7 +404,7 @@ xgb_by_tree <- function(tree, object, new_data, type, ...) {
404 404
} else {
405 405
if (type == "class") {
406 406
pred <- object$spec$method$pred$class$post(pred, object)
407 -
pred <- tibble(.pred = factor(pred, levels = object$lvl))
407 +
pred <- tibble(.pred_class = factor(pred, levels = object$lvl))
408 408
} else {
409 409
pred <- object$spec$method$pred$prob$post(pred, object)
410 410
pred <- as_tibble(pred)
@@ -503,7 +503,7 @@ C50_by_tree <- function(tree, object, new_data, type, ...) {
503 503 504 504
# switch based on prediction type
505 505
if (type == "class") {
506 -
pred <- tibble(.pred = factor(pred, levels = object$lvl))
506 +
pred <- tibble(.pred_class = factor(pred, levels = object$lvl))
507 507
} else {
508 508
pred <- as_tibble(pred)
509 509
names(pred) <- paste0(".pred_", names(pred))
@@ -309,7 +309,7 @@ multi_predict._lognet <-
309 309
if (is.null(type))
310 310
type <- "class"
311 311
if (!(type %in% c("class", "prob", "link", "raw"))) {
312 -
stop ("`type` should be either 'class', 'link', 'raw', or 'prob'.", call. = FALSE)
312 +
stop("`type` should be either 'class', 'link', 'raw', or 'prob'.", call. = FALSE)
313 313
}
314 314
if (type == "prob")
315 315
dots$type <- "response"
@@ -321,12 +321,12 @@ multi_predict._lognet <-
321 321
param_key <- tibble(group = colnames(pred), penalty = penalty)
322 322
pred <- as_tibble(pred)
323 323
pred$.row <- 1:nrow(pred)
324 -
pred <- gather(pred, group, .pred, -.row)
324 +
pred <- gather(pred, group, .pred_class, -.row)
325 325
if (dots$type == "class") {
326 -
pred[[".pred"]] <- factor(pred[[".pred"]], levels = object$lvl)
326 +
pred[[".pred_class"]] <- factor(pred[[".pred_class"]], levels = object$lvl)
327 327
} else {
328 328
if (dots$type == "response") {
329 -
pred[[".pred2"]] <- 1 - pred[[".pred"]]
329 +
pred[[".pred2"]] <- 1 - pred[[".pred_class"]]
330 330
names(pred) <- c(".row", "group", paste0(".pred_", rev(object$lvl)))
331 331
pred <- pred[, c(".row", "group", paste0(".pred_", object$lvl))]
332 332
}
@@ -371,3 +371,4 @@ predict_raw._lognet <- function(object, new_data, opts = list(), ...) {
371 371
object$spec <- eval_args(object$spec)
372 372
predict_raw.model_fit(object, new_data = new_data, opts = opts, ...)
373 373
}
374 + Original file line number Diff line number Diff line change
@@ -290,7 +290,7 @@ multi_predict._multnet <-
290 290
pred <-
291 291
tibble(
292 292
.row = rep(1:nrow(new_data), length(penalty)),
293 -
.pred = as.vector(pred),
293 +
.pred_class = as.vector(pred),
294 294
penalty = rep(penalty, each = nrow(new_data))
295 295
)
296 296
}
@@ -119,7 +119,7 @@ test_that('glmnet prediction, mulitiple lambda', {
119 119
mult_pred$rows <- rep(1:7, 2)
120 120
mult_pred <- mult_pred[order(mult_pred$rows, mult_pred$penalty), ]
121 121
mult_pred <- mult_pred[, c("penalty", "values")]
122 -
names(mult_pred) <- c("penalty", ".pred")
122 +
names(mult_pred) <- c("penalty", ".pred_class")
123 123
mult_pred <- tibble::as_tibble(mult_pred)
124 124 125 125
expect_equal(
@@ -148,7 +148,7 @@ test_that('glmnet prediction, mulitiple lambda', {
148 148
form_pred$rows <- rep(1:7, 2)
149 149
form_pred <- form_pred[order(form_pred$rows, form_pred$penalty), ]
150 150
form_pred <- form_pred[, c("penalty", "values")]
151 -
names(form_pred) <- c("penalty", ".pred")
151 +
names(form_pred) <- c("penalty", ".pred_class")
152 152
form_pred <- tibble::as_tibble(form_pred)
153 153 154 154
expect_equal(
@@ -180,7 +180,7 @@ test_that('glmnet prediction, no lambda', {
180 180
mult_pred$rows <- rep(1:7, 2)
181 181
mult_pred <- mult_pred[order(mult_pred$rows, mult_pred$penalty), ]
182 182
mult_pred <- mult_pred[, c("penalty", "values")]
183 -
names(mult_pred) <- c("penalty", ".pred")
183 +
names(mult_pred) <- c("penalty", ".pred_class")
184 184
mult_pred <- tibble::as_tibble(mult_pred)
185 185 186 186
expect_equal(mult_pred, multi_predict(xy_fit, lending_club[1:7, num_pred]) %>% unnest())
@@ -206,7 +206,7 @@ test_that('glmnet prediction, no lambda', {
206 206
form_pred$rows <- rep(1:7, 2)
207 207
form_pred <- form_pred[order(form_pred$rows, form_pred$penalty), ]
208 208
form_pred <- form_pred[, c("penalty", "values")]
209 -
names(form_pred) <- c("penalty", ".pred")
209 +
names(form_pred) <- c("penalty", ".pred_class")
210 210
form_pred <- tibble::as_tibble(form_pred)
211 211 212 212
expect_equal(
@@ -123,7 +123,7 @@ test_that('glmnet probabilities, mulitiple lambda', {
123 123 124 124
mult_class <- names(mult_probs)[apply(mult_probs, 1, which.max)]
125 125
mult_class <- tibble(
126 -
.pred = mult_class,
126 +
.pred_class = mult_class,
127 127
penalty = rep(lams, each = 3),
128 128
row = rep(1:3, 2)
129 129
)
You can’t perform that action at this time.
RetroSearch is an open source project built by @garambo | Open a GitHub Issue
Search and Browse the WWW like it's 1997 | Search results from DuckDuckGo
HTML:
3.2
| Encoding:
UTF-8
| Version:
0.7.4