| Type: | Package |
| Title: | Time Series Forecasting Using Nearest Neighbors |
| Version: | 0.6.0 |
| Description: | Allows forecasting time series using nearest neighbors regression Francisco Martinez, Maria P. Frias, Maria D. Perez-Godoy and Antonio J. Rivera (2019) <doi:10.1007/s10462-017-9593-z>. When the forecasting horizon is higher than 1, two multi-step ahead forecasting strategies can be used. The model built is autoregressive, that is, it is only based on the observations of the time series. The nearest neighbors used in a prediction can be consulted and plotted. |
| Maintainer: | Francisco Martinez <fmartin@ujaen.es> |
| License: | GPL-2 |
| Encoding: | UTF-8 |
| RoxygenNote: | 7.2.3 |
| Depends: | R (≥ 3.6.0) |
| Suggests: | knitr, rmarkdown, testthat (≥ 3.0.0) |
| Imports: | ggplot2 (≥ 3.1.1), graphics, Rcpp, stats, utils |
| VignetteBuilder: | knitr |
| URL: | https://github.com/franciscomartinezdelrio/tsfknn |
| BugReports: | https://github.com/franciscomartinezdelrio/tsfknn/issues |
| LinkingTo: | Rcpp |
| Config/testthat/edition: | 3 |
| NeedsCompilation: | yes |
| Packaged: | 2023-12-20 16:08:59 UTC; UJA |
| Author: | Francisco Martinez [aut, cre] |
| Repository: | CRAN |
| Date/Publication: | 2023-12-20 16:20:02 UTC |
tsfknn: Time Series Forecasting Using Nearest Neighbors
Description
Allows forecasting time series using nearest neighbors regression Francisco Martinez, Maria P. Frias, Maria D. Perez-Godoy and Antonio J. Rivera (2019) doi:10.1007/s10462-017-9593-z. When the forecasting horizon is higher than 1, two multi-step ahead forecasting strategies can be used. The model built is autoregressive, that is, it is only based on the observations of the time series. The nearest neighbors used in a prediction can be consulted and plotted.
Author(s)
Maintainer: Francisco Martinez fmartin@ujaen.es
See Also
Useful links:
Report bugs at https://github.com/franciscomartinezdelrio/tsfknn/issues
Create a ggplot object from a knnForecast object
Description
It uses a knnForecast object to create a ggplot object that plots a time series and its forecast using KNN regression.
Usage
## S3 method for class 'knnForecast'
autoplot(object, ...)
Arguments
object |
An object of class |
... |
additional parameter, see details. |
Details
Commonly used parameters are:
-
highlight. A character string indicating what elements should be highlighted. Possible values are"none","points"and"neighbors". The default value is"none". -
faceting. Logical. This applies only if thehighlightparameter is set to"neighbors". It indicates whether the different nearest neighbors should be seen in different plots (TRUE, the default value) or in one plot.
Value
The ggplot object representing a plotting with the forecast.
Examples
pred <- knn_forecasting(USAccDeaths, h = 12, lags = 1:12, k = 2)
library(ggplot2)
autoplot(pred)
autoplot(pred, highlight = "neighbors")
Examples of the model associated with a prediction
Description
It allows to see the examples of the model associated to a
knnForecast object.
Usage
knn_examples(forecast)
Arguments
forecast |
A |
Value
A matrix including the features and targets of the examples
associated with the model of a knnForecast object.
Examples
pred <- knn_forecasting(ts(1:8), h = 1, lags = 1:2, k = 2)
knn_examples(pred)
Time series forecasting using KNN regression
Description
It applies KNN regression to forecast the future values of a time series.
The lags used as autoregressive variables are set with the lags
parameter. If the user does not set the number of nearest neighbors or
the lags, these values are selected automatically.
Usage
knn_forecasting(
timeS,
h,
lags = NULL,
k = c(3, 5, 7),
msas = c("recursive", "MIMO"),
cf = c("mean", "median", "weighted"),
transform = c("additive", "multiplicative", "none")
)
Arguments
timeS |
A numeric vector or time series of class |
h |
A positive integer. Number of values to forecast. |
lags |
An integer vector in increasing order expressing the lags used as autoregressive variables. |
k |
A positive integer. The k parameter in KNN regression. A vector of k values can also be used. In that case, the forecast is the average of the forecasts produced by the different models with the different k parameters. |
msas |
A string indicating the Multiple-Step Ahead Strategy used when more than one value is predicted. It can be "recursive" or "MIMO" (the default). |
cf |
A string. It indicates the combination function used to aggregate the targets associated with the nearest neighbors. It can be "median", "weighted" or "mean" (the default). |
transform |
A character value indicating whether the training samples
are transformed. If the time series has a trend it is recommended. By
default is |
Value
An object of class "knnForecast". The
function summary can be used to obtain or print a
summary of the results.
An object of class \code{"knnForecast"} is a list containing at least
the following components:
call |
the matched call. |
msas |
the Multi-Step Ahead Strategy. |
prediction |
a time series with the forecast. |
model |
an object of class |
Examples
pred <- knn_forecasting(USAccDeaths, h = 12, lags = 1:12, k = 2)
pred$prediction # To see a time series with the forecasts
plot(pred) # To see a plot with the forecast
Number of training examples
Description
It computes the number of training examples that would have a KNN model with the specified parameters.
Usage
n_training_examples(timeS, h, lags, msas = c("MIMO", "recursive"))
Arguments
timeS |
A numeric vector or time series of class |
h |
A positive integer. Number of values to forecast. |
lags |
An integer vector in increasing order expressing the lags used as autoregressive variables. |
msas |
A string indicating the Multiple-Step Ahead Strategy used when more than one value is predicted. It can be "recursive" or "MIMO" (the default). |
Value
An integer.
Examples
n_training_examples(ts(1:10), h = 2, lags = 1:3, msas = "MIMO")
n_training_examples(ts(1:10), h = 2, lags = 1:3, msas = "recursive")
Nearest neighbors associated with predictions
Description
It allows to check the new instances and their nearest neighbors used in a prediction associated with a "knnForecast" object.
Usage
nearest_neighbors(forecast)
Arguments
forecast |
A |
Value
A list including the new instances used in KNN regression and their nearest neighbors.
Examples
pred <- knn_forecasting(UKgas, h = 4, lags = 1:4, k = 2, msas = "MIMO")
nearest_neighbors(pred)
Plot a prediction of a test set
Description
It uses a test set generated with the function rolling_origin
and plots its forecast.
Usage
## S3 method for class 'knnForecastRO'
plot(x, h = NULL, ...)
Arguments
x |
the object obtained from a call to |
h |
an integer. The forecasting horizon. If |
... |
Other plotting parameters to affect the plot. |
Predict method for KNN models for time series forecasting.
Description
Predicted values based on a KNN model for time series forecasting.
Usage
## S3 method for class 'knnForecast'
predict(object, h, ...)
Arguments
object |
a |
h |
an integer. The forecasting horizon. |
... |
further arguments passed to or from other methods. |
Details
If the models uses the MIMO strategy for multiple-step ahead prediction, the forecasting horizon is fixed to the model forecasting horizon.
Value
a knnForecast object with the prediction and information
about the KNN model, see the documentation of knn_forecasting
for the structure of knnForecast objects.
Examples
pred <- knn_forecasting(UKgas, h = 4, k = 1, msas = "recursive")
new_pred <- predict(pred, h = 6)
print(new_pred$prediction)
plot(new_pred) # To see a plot with the forecast
Assessing forecasting accuracy with rolling origin
Description
It uses the model and the time series associated with the knnForecast
object to asses the forecasting accuracy of the model using the last
h values of the time series to build test sets applying a rolling
origin evaluation.
Usage
rolling_origin(knnf, h = NULL, rolling = TRUE)
Arguments
knnf |
A |
h |
A positive integer. The forecast horizon. If |
rolling |
A logical. If |
Details
This function assesses the forecast accuracy of the model used by the
knnForecast object. It uses h different test and training
sets. The first test set consists of the last h values of the time
series (the training set is formed by the previous values). The next test
set consists of the last h - 1 values of the time series and so on
(the last test set is formed by the last value of the time series).
Value
A list containing at least the following fields:
test_sets |
a matrix containing the test sets used in the evaluation. Every row contains a different test set. |
predictions |
The predictions for the test sets. |
errors |
The errors for the test sets. |
global_accu |
Different measures of accuracy applied to all the errors. |
h_accu |
Different measures of accuracy applied to all the errors for every forecasting horizon. |
Examples
pred <- knn_forecasting(UKgas, h = 4, lags = 1:4, k = 2)
ro <- rolling_origin(pred)
print(ro$global_accu)