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
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
|
#' Simple Value Assignments for Novel Factor Levels
#'
#' `step_novel` creates a *specification* of a recipe
#' step that will assign a previously unseen factor level to a
#' new value.
#'
#' @inheritParams step_center
#' @inherit step_center return
#' @param ... One or more selector functions to choose which
#' variables that will be affected by the step. These variables
#' should be character or factor types. See [selections()] for more
#' details. For the `tidy` method, these are not currently used.
#' @param role Not used by this step since no new variables are
#' created.
#' @param new_level A single character value that will be assigned
#' to new factor levels.
#' @param objects A list of objects that contain the information
#' on factor levels that will be determined by [prep.recipe()].
#' @return An updated version of `recipe` with the new step
#' added to the sequence of existing steps (if any). For the
#' `tidy` method, a tibble with columns `terms` (the
#' columns that will be affected) and `value` (the factor
#' levels that is used for the new value)
#' @keywords datagen
#' @concept preprocessing
#' @concept factors
#' @export
#' @details The selected variables are adjusted to have a new
#' level (given by `new_level`) that is placed in the last
#' position. During preparation there will be no data points
#' associated with this new level since all of the data have been
#' seen.
#'
#' Note that if the original columns are character, they will be
#' converted to factors by this step.
#'
#' Missing values will remain missing.
#'
#' If `new_level` is already in the data given to `prep`, an error
#' is thrown.
#'
#' @seealso [step_factor2string()], [step_string2factor()],
#' [dummy_names()], [step_regex()], [step_count()],
#' [step_ordinalscore()], [step_unorder()], [step_other()]
#' @examples
#' library(modeldata)
#' data(okc)
#'
#' okc_tr <- okc[1:30000,]
#' okc_te <- okc[30001:30006,]
#' okc_te$diet[3] <- "cannibalism"
#' okc_te$diet[4] <- "vampirism"
#'
#' rec <- recipe(~ diet + location, data = okc_tr)
#'
#' rec <- rec %>%
#' step_novel(diet, location)
#' rec <- prep(rec, training = okc_tr)
#'
#' processed <- bake(rec, okc_te)
#' tibble(old = okc_te$diet, new = processed$diet)
#'
#' tidy(rec, number = 1)
step_novel <-
function(recipe,
...,
role = NA,
trained = FALSE,
new_level = "new",
objects = NULL,
skip = FALSE,
id = rand_id("novel")) {
add_step(
recipe,
step_novel_new(
terms = ellipse_check(...),
role = role,
trained = trained,
new_level = new_level,
objects = objects,
skip = skip,
id = id
)
)
}
step_novel_new <-
function(terms, role, trained, new_level, objects, skip, id) {
step(
subclass = "novel",
terms = terms,
role = role,
trained = trained,
new_level = new_level,
objects = objects,
skip = skip,
id = id
)
}
get_existing_values <- function(x) {
if (is.character(x)) {
out <- unique(x)
attr(out, "is_ordered") <- FALSE
} else {
if (is.factor(x)) {
out <- levels(x)
attr(out, "is_ordered") <- is.ordered(x)
}
else
rlang::abort("Data should be either character or factor")
}
out
}
#' @export
prep.step_novel <- function(x, training, info = NULL, ...) {
col_names <- eval_select_recipes(x$terms, training, info)
col_check <- dplyr::filter(info, variable %in% col_names)
if (any(col_check$type != "nominal"))
rlang::abort(
paste0("Columns must be character or factor: ",
paste0(col_check$variable[col_check$type != "nominal"],
collapse = ", "))
)
# Get existing levels and their factor type (i.e. ordered)
objects <- lapply(training[, col_names], get_existing_values)
# Check to make sure that there are not duplicate levels
level_check <-
map_lgl(objects, function(x, y) y %in% x, y = x$new_level)
if (any(level_check))
rlang::abort(
paste0(
"Columns already contain the new level: ",
paste0(names(level_check)[level_check], collapse = ", ")
)
)
step_novel_new(
terms = x$terms,
role = x$role,
trained = TRUE,
new_level = x$new_level,
objects = objects,
skip = x$skip,
id = x$id
)
}
#' @export
bake.step_novel <- function(object, new_data, ...) {
for (i in names(object$objects)) {
new_data[[i]] <- ifelse(
# Preserve NA values by adding them to the list of existing
# possible values
!(new_data[[i]] %in% c(object$object[[i]], NA)),
object$new_level,
as.character(new_data[[i]])
)
new_data[[i]] <-
factor(new_data[[i]],
levels = c(object$object[[i]], object$new_level),
ordered = attributes(object$object[[i]])$is_ordered)
}
if (!is_tibble(new_data))
new_data <- as_tibble(new_data)
new_data
}
print.step_novel <-
function(x, width = max(20, options()$width - 30), ...) {
cat("Novel factor level assignment for ", sep = "")
printer(names(x$objects), x$terms, x$trained, width = width)
invisible(x)
}
#' @rdname step_novel
#' @param x A `step_novel` object.
#' @export
tidy.step_novel <- function(x, ...) {
if (is_trained(x)) {
res <- tibble(terms = names(x$objects),
value = rep(x$new_level, length(x$objects)))
} else {
term_names <- sel2char(x$terms)
res <- tibble(terms = term_names,
value = rep(x$new_level, length(term_names)))
}
res$id <- x$id
res
}
|
