univariate time series into training and test splitting in R

Question

I want to split the univariate time series (14139 observations) into training and test set for 60% and 40% respectively. I enter the command
splits (APILts, c(rep("train", 8483), "test")) then R resulted: Error: is.timeSeries(x) is not TRUE

Answer 1

The forecast package by Rob Hyndman supports since version 8.0 head and tail for time series. This enables a very handy mechanism for creating train and test sets without the window or the subset function. All you have to do is provide the size of the training set, as exemplified in the code below.

library(forecast)
train <- head(AirPassengers, round(length(AirPassengers) * 0.6))
h <- length(AirPassengers) - length(train)
test <- tail(AirPassengers, h)

Alternatively, you can provide the length of the forecast horizon:

h2 <- 58L
train2 <- head(AirPassengers, round(length(AirPassengers) - h2))
test2 <- tail(AirPassengers, h2)

Test that they are identical:

identical(train, train2)
identical(test, test2)
identical(h, h2)
[1] TRUE
[1] TRUE
[1] TRUE

Inspect train:

train
     Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
1949 112 118 132 129 121 135 148 148 136 119 104 118
1950 115 126 141 135 125 149 170 170 158 133 114 140
1951 145 150 178 163 172 178 199 199 184 162 146 166
1952 171 180 193 181 183 218 230 242 209 191 172 194
1953 196 196 236 235 229 243 264 272 237 211 180 201
1954 204 188 235 227 234 264 302 293 259 229 203 229
1955 242 233 267 269 270 315 364 347 312 274 237 278
1956 284 277 

Inspect test:

test
     Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
1956         317 313 318 374 413 405 355 306 271 306
1957 315 301 356 348 355 422 465 467 404 347 305 336
1958 340 318 362 348 363 435 491 505 404 359 310 337
1959 360 342 406 396 420 472 548 559 463 407 362 405
1960 417 391 419 461 472 535 622 606 508 461 390 432

Plot train and test:

autoplot(train) + autolayer(test)

Answer 2

To split a time series you need a vector that is a time series.

The error suggest that your APILts is not a ts object:

Error: is.timeSeries(x) is not TRUE

Here an example on how to split a time series with a ts object:

data(AirPassengers)

Data are assigned to a convenient vector

This is a easy way to avoid changing the code every time

series <- AirPassengers

Plot the series

plot(series, col="darkblue", ylab="Passegners on airplanes")

Plot the seasonal distribution of the series

windows(width=800,height=350)   # set the window with the dimensions you need

boxplot(split(series, cycle(series)), names = month.abb, col = "gold")

The size of the test set is typically about 40% of the total sample

So, we will split the series in a training set and a test set

# Training set
# Use data from 1949 to 1955 for forecasting

sr = window(series, start=1949, end=c(1955,12))

# Test set
# Use remaining data from 1956 to 1960 to test accuracy

ser = window(series, start=1956, end=c(1960,12))

Now we are ready to start.

To convert data into time series

# Data 

dat <- c(27, 28, 25, 22, 19, 21, 24, 24, 22, 16, 27, 41, 29, 24, 15, 27, 25, 21, 15, 41, 19, 24, 34, 20, 25, 34, 31, 29, 38, 36, 27, 37, 31, 28, 25, 34, 40, 36, 39, 19, 40, 31, 29, 39, 29, 40, 34, 31)


# Convert the data to time series

series <- ts(dat, frequency = 12, start = c(1969, 1))

# Inspect the series

series

plot(series)

Note that sometime, if you have uploaded the data as a dataframe, you must specify the column with precision:

# In this example, the data you want to convert into a ts object are in the first column

series <- ts(dat[[1]], frequency = 12, start = c(1969, 1))