Title: Visualizing Collections of Time Series Forecasts
Version: 0.1.1
Description: A way of visualizing collections of time series and, optionally their future values, forecasts for their future values and prediction intervals for the forecasts. A web-based GUI can be used to display the information in a collection of time series.
Maintainer: Francisco Martinez <fmartin@ujaen.es>
License: MIT + file LICENSE
Encoding: UTF-8
RoxygenNote: 7.2.3
Suggests: forecast, knitr, Mcomp, rmarkdown, testthat (≥ 3.0.0)
Config/testthat/edition: 3
Imports: ggplot2, methods, shiny
URL: https://github.com/franciscomartinezdelrio/vctsfr
BugReports: https://github.com/franciscomartinezdelrio/vctsfr/issues
VignetteBuilder: knitr
NeedsCompilation: no
Packaged: 2023-12-07 18:35:18 UTC; UJA
Author: Maria Pilar Frias-Bustamante ORCID iD [aut], Francisco Martinez ORCID iD [aut, cre, cph]
Repository: CRAN
Date/Publication: 2023-12-07 18:50:02 UTC

Launches the web-based GUI for visualizing time series

Description

Launches the web-based GUI for visualizing a collection of time series in a web browser.

Usage

GUI_collection(collection)

Arguments

collection

a list with the collection of time series. Each component of the list must have been built with the ts_info() function.

Details

The vctsfr package provides a Shiny-based GUI to visualize collections of time series and their forecasts. The main features of the GUI are:

Value

Nothing

Examples


# create a collection of two time series and visualize them
c <- list(ts_info(USAccDeaths), ts_info(ldeaths))
GUI_collection(c)

Check that a collection of time series is properly formatted

Description

This function checks that an object holding a collection of time series, their future values and their forecasts has the correct format. This kind of objects are used in function plot_collection(). A collection of time series should be a list compounded of objects of class ts_info, which are built using the ts_info() function.

Usage

check_time_series_collection(collection)

Arguments

collection

a list representing a collection of time series as described in plot_collection().

Value

a character string with value "OK" if the object is properly formatted. Otherwise, the character string indicates the first error found in the object's format.

Examples

c <- list(ts_info(USAccDeaths), ts_info(ldeaths))
check_time_series_collection(c)

Create a prediction interval object

Description

The object created represents a prediction interval for the forecast of the future values of a time series.

Usage

pi_info(level, lpi, upi)

Arguments

level

a number in the interval (0, 100) indicating the level of the prediction interval.

lpi

a time series of class ts or a vector. Lower limit of a prediction interval.

upi

a time series of class ts or a vector. Upper limit of a prediction interval.

Value

An object of class pi_info. It is a list containing all the information supplied to the function.

See Also

prediction_info() which uses this function to specify prediction intervals.

Examples

if (require("forecast")) {
  time_series <- ts(rnorm(40))
  f <- meanf(time_series, level = 95)
  info <- pi_info(95, f$lower, f$upper)
}

Create a ggplot object associated with a time series belonging to a collection.

Description

Apart from the time series, future values and forecasts for the future values form part of the ggplot object.

Usage

plot_collection(collection, number, methods = NULL, level = NULL, sdp = TRUE)

Arguments

collection

a list with the collection of time series. Each component of the list must have been built with the ts_info() function.

number

an integer. The number of the time series. It should be a value between 1 and length(collection).

methods

NULL (default) or a character vector indicating the names of the forecasting methods to be displayed.

level

NULL (default) or a number in the interval (0, 100) indicating the level of the prediction interval to be shown. This parameter in considered only when just one forecasting method is plotted and the forecasting method has a prediction interval with the specified level.

sdp

logical. Should data points be shown in the plot? (default value TRUE)

Details

The collection parameter must be a list. Each component of the list stores a time series and, optionally, its future values, forecasts for the future values and prediction intervals for the forecasts. Each component should have been created using the ts_info() function.

In the example section you can see an example of a collection of time series. If the collection parameter is not specified correctly, a proper message is shown.

Value

The ggplot object representing the time series and its forecast.

See Also

ts_info() function to see how to build the components of the collection parameter.

Examples

# create a collection of two time series and plot both time series
c <- list(ts_info(USAccDeaths), ts_info(ldeaths))
plot_collection(c, number = 1)
plot_collection(c, number = 2, sdp = FALSE)

# create a collection of one time series with future values and forecasts
if (require(forecast)) {
  c <- vector(2, mode = "list")
  timeS <- window(USAccDeaths, end = c(1977, 12))
  f <- window(USAccDeaths, start = c(1978, 1))
  ets_fit   <- ets(timeS)
  ets_pred  <- forecast(ets_fit, h = length(f), level = 90)
  mean_pred <- meanf(timeS, h = length(f), level = 90)
  c[[1]] <- ts_info(timeS, future = f,
            prediction_info("ES", ets_pred$mean,
                            pi_info(90, ets_pred$lower, ets_pred$upper)),
            prediction_info("Mean", mean_pred$mean,
                            pi_info(90, mean_pred$lower, mean_pred$upper))
  )
  timeS <- ts(rnorm(30, sd = 3))
  f <- rnorm(5, sd = 3)
  rw <- rwf(timeS, h = length(f), level = 80)
  mean <- meanf(timeS, h = length(f), level = 90)
  c[[2]] <- ts_info(timeS, future = f,
            prediction_info("Random Walk", rw$mean,
                            pi_info(80, rw$lower, rw$upper)),
            prediction_info("Mean", mean$mean,
                            pi_info(90, mean$lower, mean$upper))
  )
  plot_collection(c, number = 1)
}
if (require("forecast"))
  plot_collection(c, number = 2)
if (require("forecast"))
  plot_collection(c, number = 2, methods = "Mean") # just plot a forecasting method
if (require("forecast"))
  plot_collection(c, number = 2, methods = "Random Walk", level = 80)

Creates a ggplot object with a time series and some forecasts

Description

Create a ggplot object with a time series and, optionally, some future values of the time series and several forecast for those future values.

Usage

plot_predictions(ts, future = NULL, predictions = NULL, sdp = TRUE)

Arguments

ts

a time series of class ts.

future

NULL (default) or a time series of class ts or a vector. Future values of the time series.

predictions

NULL (default) or a named list containing the predictions for the future values. Each component of the list should contain a vector or an object of class ts representing a forecast, the name of the component should be the name of the forecasting method.

sdp

logical. Should data points be shown? (default value TRUE)

Details

If future or the forecasts in the prediction list are vectors then they are supposed to start after the last data of the time series.

Value

The ggplot object representing the time series and its forecast.

Examples

# plot a time series, its future values and two forecasts
ts <- window(USAccDeaths, end = c(1977, 12))
f <- window(USAccDeaths, start = c(1978, 1))
prediction1 <- rep(mean(ts), 12)
prediction2 <- as.vector(window(ts, start = c(1977, 1)))
p <- list(Mean = prediction1, Naive = prediction2)
plot_predictions(ts, future = f, predictions = p)

Create a ggplot object with a time series and forecast

Description

Create a ggplot object associated with a time series and, optionally, its future values, a forecast for its future values and a prediction interval of the forecast.

Usage

plot_ts(
  ts,
  future = NULL,
  prediction = NULL,
  method = NULL,
  lpi = NULL,
  upi = NULL,
  level = NULL,
  sdp = TRUE
)

Arguments

ts

a time series of class ts.

future

NULL (default) or a time series of class ts or a vector. Future values of the time series.

prediction

NULL (default) or a time series of class ts or a vector. Forecast of the future values of the time series.

method

NULL (default) a character string with the name of the method used to forecast the future values of the time series. This name will appear in the legend.

lpi

NULL (default) or a time series of class ts or a vector. Lower limit of a prediction interval for the prediction parameter.

upi

NULL (default) or a time series of class ts or a vector. Upper limit of a prediction interval for the prediction parameter.

level

NULL (default) a number in the interval (0, 100) indicating the level of the prediction interval.

sdp

logical. Should data points be shown? (default value TRUE)

Details

If future or prediction are vectors then they are supposed to start after the last data of the time series.

Value

The ggplot object representing the time series and its forecast.

Examples

library(ggplot2)
plot_ts(USAccDeaths) # plot a time series

# plot a time series, not showing data points
plot_ts(USAccDeaths, sdp = FALSE)

# plot a time series, its future values and a prediction
ts <- window(USAccDeaths, end = c(1977, 12))
f <- window(USAccDeaths, start = c(1978, 1))
p <- ts(window(USAccDeaths, start = c(1976, 1), end = c(1976, 12)),
        start = c(1978, 1),
        frequency = 12
)
plot_ts(ts, future = f, prediction = p)

# plot a time series and a prediction
plot_ts(USAccDeaths, prediction = rep(mean(USAccDeaths), 12),
        method = "Mean")

# plot a time series, a prediction and a prediction interval
if (require(forecast)) {
  timeS <- window(USAccDeaths, end = c(1977, 12))
  f <- window(USAccDeaths, start = c(1978, 1))
  ets_fit <- ets(timeS)
  p <- forecast(ets_fit, h = length(f), level = 90)
  plot_ts(timeS, future = f, prediction = p$mean, method = "ES",
          lpi = p$lower, upi = p$upper, level = 90
  )
}

Create an object with a prediction about the future values of a time series

Description

The object created contains a forecast and, optionally, prediction intervals for the forecast.

Usage

prediction_info(name, forecast, ...)

Arguments

name

a character indicating the name of the method used to forecast.

forecast

a time series of class ts or a vector. It is a prediction for the future values of a time series.

...

prediction intervals for the forecast. These prediction intervals must have been built with the pi_info() function.

Value

an object of class pred_info. A list with the information supplied to the function.

See Also

pi_info() for how to create prediction intervals.

Examples

if (require("forecast")) {
  time_series <- ts(rnorm(40))
  f <- meanf(time_series, level = 95)
  info <- prediction_info("mean", f$mean, pi_info(95, f$lower, f$upper))
}

Create an object with information about a time series

Description

The information about the time series is compounded of the time series and, optionally, its future values and forecasts for those future values (and prediction intervals for those forecasts).

Usage

ts_info(historical, ..., future = NULL, name = NULL)

Arguments

historical

a time series of class ts with the historical values of the series.

...

forecasts for the future values of the time series. A forecast must have been built with the prediction_info() function. See the examples section.

future

NULL (default) or a time series of class ts or a vector. The future values of the time series (possibly to be forecast).

name

NULL (default) or a character string with information about the time series. Typically, its name.

Value

An object of class ts_info. It is a list containing all the information supplied to the function.

See Also

prediction_info() for how to create forecasts.

Examples

# only information about a time series
info <- ts_info(USAccDeaths)

# Information about a time series and its future values
info2 <- ts_info(ts(rnorm(50)), future = rnorm(10))

# Information about a time series, its future values and a forecast
if (require("forecast")) {
  t <- ts(rnorm(50))
  f <- rnorm(10)
  mf <- meanf(t, level = 95)
  info3 <- ts_info(t, future = f,
                   prediction_info("mean", mf$mean,
                                   pi_info(95, mf$lower, mf$upper)
                   )
  )
}