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
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
|
#' @importFrom stats as.formula
#' @export
densityplot.train <- function(x,
data = NULL,
metric = x$metric,
...)
{
if (!is.null(match.call()$data))
warning("explicit 'data' specification ignored")
if(x$control$method %in% c("oob", "LOOCV"))
stop("Resampling plots cannot be done with leave-out-out CV or out-of-bag resampling")
resamp <- x$resample
tNames <- gsub("^\\.", "", names(x$bestTune))
# adapt formula to work with muliple metrics
mName <- names(resamp)[names(resamp) %in% metric][1]
# Look for constant tuning parameters and remove them
numVals <- unlist(
lapply(resamp,
function(u) length(unique(u))))
if(any(numVals == 1))
{
# make sure that these are tuning parameters
resamp <- resamp[, numVals > 1, drop = FALSE]
tNames <- tNames[tNames %in% names(numVals)[numVals > 1]]
}
# Create the formula based on the data
formText <- paste("~", mName)
if(any(tNames %in% colnames(resamp)))
{
formText <- paste(formText,
"|",
paste(
tNames,
collapse = "*"))
}
form <- as.formula(formText)
densityplot(form, data = resamp, ...)
}
#' Lattice functions for plotting resampling results
#'
#' A set of lattice functions are provided to plot the resampled performance
#' estimates (e.g. classification accuracy, RMSE) over tuning parameters (if
#' any).
#'
#' By default, only the resampling results for the optimal model are saved in
#' the \code{train} object. The function \code{\link{trainControl}} can be used
#' to save all the results (see the example below).
#'
#' If leave-one-out or out-of-bag resampling was specified, plots cannot be
#' produced (see the \code{method} argument of \code{\link{trainControl}})
#'
#' For \code{xyplot} and \code{stripplot}, the tuning parameter with the most
#' unique values will be plotted on the x-axis. The remaining parameters (if
#' any) will be used as conditioning variables. For \code{densityplot} and
#' \code{histogram}, all tuning parameters are used for conditioning.
#'
#' Using \code{horizontal = FALSE} in \code{stripplot} works.
#'
#' @aliases stripplot.train xyplot.train densityplot.train histogram.train
#' @param x An object produced by \code{\link{train}}
#' @param data This argument is not used
#' @param metric A character string specifying the single performance metric
#' that will be plotted
#' @param \dots arguments to pass to either
#' \code{\link[lattice:histogram]{histogram}},
#' \code{\link[lattice:histogram]{densityplot}},
#' \code{\link[lattice:xyplot]{xyplot}} or
#' \code{\link[lattice:xyplot]{stripplot}}
#' @return A lattice plot object
#' @author Max Kuhn
#' @seealso \code{\link{train}}, \code{\link{trainControl}},
#' \code{\link[lattice:histogram]{histogram}},
#' \code{\link[lattice:histogram]{densityplot}},
#' \code{\link[lattice:xyplot]{xyplot}},
#' \code{\link[lattice:xyplot]{stripplot}}
#' @keywords hplot
#' @examples
#'
#' \dontrun{
#'
#' library(mlbench)
#' data(BostonHousing)
#'
#' library(rpart)
#' rpartFit <- train(medv ~ .,
#' data = BostonHousing,
#' "rpart",
#' tuneLength = 9,
#' trControl = trainControl(
#' method = "boot",
#' returnResamp = "all"))
#'
#' densityplot(rpartFit,
#' adjust = 1.25)
#'
#' xyplot(rpartFit,
#' metric = "Rsquared",
#' type = c("p", "a"))
#'
#' stripplot(rpartFit,
#' horizontal = FALSE,
#' jitter = TRUE)
#'
#' }
#' @export
histogram.train <- function(x,
data = NULL,
metric = x$metric,
...)
{
if (!is.null(match.call()$data))
warning("explicit 'data' specification ignored")
if(x$control$method %in% c("oob", "LOOCV"))
stop("Resampling plots cannot be done with leave-out-out CV or out-of-bag resampling")
resamp <- x$resample
tNames <- gsub("^\\.", "", names(x$bestTune))
# adapt formula to work with muliple metrics
mName <- names(resamp)[names(resamp) %in% metric][1]
# Look for constant tuning parameters and remove them
numVals <- unlist(
lapply(resamp,
function(u) length(unique(u))))
if(any(numVals == 1))
{
# make sure that these are tuning parameters
resamp <- resamp[, numVals > 1, drop = FALSE]
tNames <- tNames[tNames %in% names(numVals)[numVals > 1]]
}
# Create the formula based on the data
formText <- paste("~", mName)
if(any(tNames %in% colnames(resamp)))
{
formText <- paste(formText,
"|",
paste(
tNames,
collapse = "*"))
}
form <- as.formula(formText)
histogram(form, data = resamp, ...)
}
#' @importFrom stats as.formula
#' @export
stripplot.train <- function(x,
data = NULL,
metric = x$metric,
...)
{
if (!is.null(match.call()$data))
warning("explicit 'data' specification ignored")
if(x$control$method %in% c("oob", "LOOCV"))
stop("Resampling plots cannot be done with leave-out-out CV or out-of-bag resampling")
resamp <- x$resample
tNames <- gsub("^\\.", "", names(x$bestTune))
# adapt formula to work with muliple metrics
mName <- names(resamp)[names(resamp) %in% metric][1]
# Look for constant tuning parameters and remove them
numVals <- unlist(
lapply(resamp,
function(u) length(unique(u))))
if(any(numVals == 1))
{
# make sure that these are tuning parameters
resamp <- resamp[, numVals > 1, drop = FALSE]
tNames <- tNames[tNames %in% names(numVals)[numVals > 1]]
}
# determine which tuning parameter has the most values
tNames1 <- names(which.max(numVals[names(numVals) %in% tNames]))
tNames2 <- tNames[!(tNames %in% tNames1)]
# The variable in tNames1 will be converted to a factor, so
# we will make sure that numeric data gets changed correctly
resamp[,tNames1] <- factor(
as.character(resamp[,tNames1]),
levels = paste(
sort(unique(resamp[,tNames1]))))
# Create the formula based on the data
if(any(tNames %in% colnames(resamp)))
{
theDots <- list(...)
if(any(names(theDots) == "horizontal"))
{
formText <- if(theDots$horizontal) paste(tNames1, "~", mName) else paste(mName, "~", tNames1)
} else formText <- paste(tNames1, "~", mName)
if(length(tNames2) > 0)
{
formText <- paste(formText,
"|",
paste(
tNames2,
collapse = "*"))
}
} else formText <- paste("~", mName)
form <- as.formula(formText)
stripplot(form, data = resamp, ...)
}
#' @importFrom stats as.formula
#' @export
xyplot.train <- function(x,
data = NULL,
metric = x$metric,
...)
{
if (!is.null(match.call()$data))
warning("explicit 'data' specification ignored")
if(x$control$method %in% c("oob", "LOOCV"))
stop("Resampling plots cannot be done with leave-out-out CV or out-of-bag resampling")
resamp <- x$resample
tNames <- gsub("^\\.", "", names(x$bestTune))
# adapt formula to work with muliple metrics
mName <- names(resamp)[names(resamp) %in% metric][1]
# Look for constant tuning parameters and remove them
numVals <- unlist(
lapply(resamp,
function(u) length(unique(u))))
if(any(numVals == 1))
{
# make sure that these are tuning parameters
resamp <- resamp[, numVals > 1, drop = FALSE]
tNames <- tNames[tNames %in% names(numVals)[numVals > 1]]
}
# determine which tuning parameter has the most values
tNames1 <- names(which.max(numVals[names(numVals) %in% tNames]))
tNames2 <- tNames[!(tNames %in% tNames1)]
# Create the formula based on the data
if(any(tNames %in% colnames(resamp)))
{
formText <- paste(mName, "~", tNames1)
if(length(tNames2) > 0)
{
formText <- paste(formText,
"|",
paste(
tNames2,
collapse = "*"))
}
} else stop("there must be at least one tuning parameter for a scatter plot")
form <- as.formula(formText)
xyplot(form, data = resamp, ...)
}
|
