Title: Conformal Time Series Forecasting Using State of Art Machine Learning Algorithms
Version: 0.1.2
Description: Conformal time series forecasting using the caret infrastructure. It provides access to state-of-the-art machine learning models for forecasting applications. The hyperparameter of each model is selected based on time series cross-validation, and forecasting is done recursively.
License: GPL (≥ 3)
URL: https://akai01.github.io/caretForecast/, https://github.com/Akai01/caretForecast
BugReports: https://github.com/Akai01/caretForecast/issues
Depends: R (≥ 3.6)
Imports: forecast (≥ 8.15), caret (≥ 6.0.88), magrittr (≥ 2.0.1), methods (≥ 4.1.1), dplyr (≥ 1.0.9), generics (≥ 0.1.3)
Suggests: Cubist (≥ 0.3.0), knitr (≥ 1.29), testthat (≥ 2.3.2)
Encoding: UTF-8
LazyData: true
RoxygenNote: 7.3.2
NeedsCompilation: no
Packaged: 2026-01-29 22:38:59 UTC; akayr
Author: Resul Akay [aut, cre]
Maintainer: Resul Akay <resulakay1@gmail.com>
Repository: CRAN
Date/Publication: 2026-01-30 06:10:38 UTC

Autoregressive forecasting using various Machine Learning models.

Description

Autoregressive forecasting using various Machine Learning models.

Usage

ARml(
  y,
  max_lag = 5,
  xreg = NULL,
  caret_method = "cubist",
  metric = "RMSE",
  pre_process = NULL,
  cv = TRUE,
  cv_horizon = 4,
  initial_window = NULL,
  fixed_window = FALSE,
  verbose = TRUE,
  seasonal = TRUE,
  K = frequency(y)/2,
  tune_grid = NULL,
  lambda = NULL,
  BoxCox_method = c("guerrero", "loglik"),
  BoxCox_lower = -1,
  BoxCox_upper = 2,
  BoxCox_biasadj = FALSE,
  BoxCox_fvar = NULL,
  allow_parallel = FALSE,
  calibrate = TRUE,
  calibration_horizon = NULL,
  n_cal_windows = NULL,
  ...
)

Arguments

y

A univariate time series object.

max_lag

Maximum value of lag.

xreg

Optional. A numerical vector or matrix of external regressors, which must have the same number of rows as y. (It should not be a data frame.).

caret_method

A string specifying which classification or regression model to use. Possible values are found using names(getModelInfo()). A list of functions can also be passed for a custom model function. See https://topepo.github.io/caret/ for details.

metric

A string that specifies what summary metric will be used to select the optimal model. See ?caret::train.

pre_process

A string vector that defines a pre-processing of the predictor data. Current possibilities are "BoxCox", "YeoJohnson", "expoTrans", "center", "scale", "range", "knnImpute", "bagImpute", "medianImpute", "pca", "ica" and "spatialSign". The default is no pre-processing. See preProcess and trainControl on the procedures and how to adjust them. Pre-processing code is only designed to work when x is a simple matrix or data frame.

cv

Logical, if cv = TRUE model selection will be done via cross-validation. If cv = FALSE user need to provide a specific model via tune_grid argument.

cv_horizon

The number of consecutive values in test set sample.

initial_window

The initial number of consecutive values in each training set sample.

fixed_window

Logical, if FALSE, all training samples start at 1.

verbose

A logical for printing a training log.

seasonal

Boolean. If seasonal = TRUE the fourier terms will be used for modeling seasonality.

K

Maximum order(s) of Fourier terms

tune_grid

A data frame with possible tuning values. The columns are named the same as the tuning parameters. Use getModelInfo to get a list of tuning parameters for each model or see https://topepo.github.io/caret/available-models.html. (NOTE: If given, this argument must be named.)

lambda

BoxCox transformation parameter. If lambda = NULL If lambda = "auto", then the transformation parameter lambda is chosen using BoxCox.lambda.

BoxCox_method

BoxCox.lambda argument. Choose method to be used in calculating lambda.

BoxCox_lower

BoxCox.lambda argument. Lower limit for possible lambda values.

BoxCox_upper

BoxCox.lambda argument. Upper limit for possible lambda values.

BoxCox_biasadj

InvBoxCox argument. Use adjusted back-transformed mean for Box-Cox transformations. If transformed data is used to produce forecasts and fitted values, a regular back transformation will result in median forecasts. If biasadj is TRUE, an adjustment will be made to produce mean forecasts and fitted values.

BoxCox_fvar

InvBoxCox argument. Optional parameter required if biasadj=TRUE. Can either be the forecast variance, or a list containing the interval level, and the corresponding upper and lower intervals.

allow_parallel

If a parallel backend is loaded and available, should the function use it?

calibrate

Logical. If TRUE, performs rolling-origin calibration to compute horizon-specific conformal prediction intervals. This produces properly calibrated intervals that widen with forecast horizon (trumpet shape). Default is TRUE.

calibration_horizon

Maximum forecast horizon for calibration. If NULL (default), uses 2 * frequency(y) for seasonal data or 10 for non-seasonal data.

n_cal_windows

Number of rolling windows for calibration. If NULL (default), automatically determined based on data length (max 50).

...

Ignored.

Value

A list class of forecast containing the following elemets

Author(s)

Resul Akay

Examples


library(caretForecast)

train_data <- window(AirPassengers, end = c(1959, 12))

test <- window(AirPassengers, start = c(1960, 1))

ARml(train_data, caret_method = "lm", max_lag = 12) -> fit

forecast(fit, h = length(test)) -> fc

autoplot(fc) + autolayer(test)

accuracy(fc, test)

Perform rolling-origin calibration with recursive forecasting

Description

This function computes horizon-specific nonconformity scores by performing rolling-origin evaluation with recursive multi-step forecasting. This ensures proper out-of-sample calibration that respects the exchangeability assumption required for valid conformal prediction intervals.

Usage

calibrate_horizon_scores(
  y,
  y_modified,
  max_lag,
  caret_method,
  seasonal,
  K,
  lambda,
  pre_process,
  tune_grid,
  xreg = NULL,
  calibration_horizon,
  n_windows = NULL,
  initial_window = NULL,
  verbose = FALSE
)

Arguments

y

Original time series (untransformed)

y_modified

Transformed time series (Box-Cox if applicable)

max_lag

Maximum lag used in the model

caret_method

The caret method name

seasonal

Logical, whether seasonal terms are used

K

Fourier order for seasonality

lambda

Box-Cox transformation parameter

pre_process

Pre-processing specification

tune_grid

Tuning grid (uses best parameters from initial fit)

xreg

External regressors matrix (optional)

calibration_horizon

Maximum forecast horizon for calibration

n_windows

Number of rolling windows for calibration

initial_window

Initial training window size for calibration

verbose

Logical, print progress

Value

A list with horizon-indexed vectors of sorted absolute errors

Fit a conformal regressor.

Description

Fit a conformal regressor.

Usage

conformalRegressor(residuals, sigmas = NULL)

Arguments

residuals

Model residuals.

sigmas

A vector of difficulty estimates

Value

A conformalRegressor object

Author(s)

Resul Akay

References

Boström, H., 2022. crepes: a Python Package for Generating Conformal Regressors and Predictive Systems. In Conformal and Probabilistic Prediction and Applications. PMLR, 179.

Fit a horizon-specific conformal regressor for time series forecasting.

Description

This function creates a conformal regressor that accounts for increasing uncertainty at longer forecast horizons. It uses separate nonconformity score distributions for each horizon h=1,2,3,..., resulting in prediction intervals that naturally widen as the forecast horizon increases (trumpet-shaped intervals).

Usage

conformalRegressorByHorizon(horizon_errors)

Arguments

horizon_errors

A named list where each element contains sorted absolute errors for that horizon. Names should be "h1", "h2", etc. This is typically produced by calibrate_horizon_scores().

Value

A conformalRegressorByHorizon object containing:

alphas_by_horizon

List of sorted nonconformity scores for each horizon

max_horizon

Maximum calibrated horizon

n_samples

Number of calibration samples per horizon

Author(s)

Resul Akay

References

Boström, H., 2022. crepes: a Python Package for Generating Conformal Regressors and Predictive Systems. In Conformal and Probabilistic Prediction and Applications. PMLR, 179.

Stankeviciute, K., Alaa, A. M., & van der Schaar, M., 2021. Conformal Time-series Forecasting. NeurIPS 2021.

Forecasting using ARml model

Description

Forecasting using ARml model

Usage

## S3 method for class 'ARml'
forecast(object, h = frequency(object$y), xreg = NULL, level = c(80, 95), ...)

Arguments

object

An object of class "ARml", the result of a call to ARml.

h

forecast horizon

xreg

Optionally, a numerical vector or matrix of future external regressors

level

Confidence level for prediction intervals.

...

Ignored

Value

A list class of forecast containing the following elemets

Author(s)

Resul Akay

Examples


library(caretForecast)

train_data <- window(AirPassengers, end = c(1959, 12))

test <- window(AirPassengers, start = c(1960, 1))

ARml(train_data, caret_method = "lm", max_lag = 12) -> fit

forecast(fit, h = length(test), level = c(80,95)) -> fc

autoplot(fc)+ autolayer(test)

accuracy(fc, test)

Internal forecast loop for calibration (without time series attributes)

Description

Internal forecast loop for calibration (without time series attributes)

Usage

forecast_loop_cal(object, xreg, h)

Variable importance for forecasting model.

Description

Variable importance for forecasting model.

Usage

get_var_imp(object, plot = TRUE)

Arguments

object

A list class of ARml or forecast object derived from ARml

plot

Boolean, if TRUE, variable importance will be ploted.

Value

A list class of "varImp.train". See varImp or a "trellis" plot.

Author(s)

Resul Akay

Examples


train <- window(AirPassengers, end = c(1959, 12))

test <- window(AirPassengers, start = c(1960, 1))

ARml(train, caret_method = "lm", max_lag = 12, trend_method = "none",
 pre_process = "center") -> fit

forecast(fit, h = length(test), level = c(80,95)) -> fc

autoplot(fc)+ autolayer(test)

accuracy(fc, test)

get_var_imp(fc, plot = TRUE)

Predict a conformalRegressor

Description

Predict a conformalRegressor

Usage

## S3 method for class 'conformalRegressor'
predict(
  object,
  y_hat = NULL,
  sigmas = NULL,
  confidence = 0.95,
  y_min = -Inf,
  y_max = Inf,
  ...
)

Arguments

object

A conformalRegressor object

y_hat

Predicted values

sigmas

Difficulty estimates

confidence

Confidence level

y_min

The minimum value to include in prediction intervals

y_max

The maximum value to include in prediction intervals

...

Ignored

Value

Prediction intervals

Author(s)

Resul Akay

Predict intervals from a horizon-specific conformal regressor

Description

This function generates prediction intervals that account for increasing uncertainty at longer forecast horizons. Each horizon h uses its own calibrated nonconformity score distribution, resulting in trumpet-shaped prediction intervals.

Usage

## S3 method for class 'conformalRegressorByHorizon'
predict(
  object,
  y_hat = NULL,
  confidence = 0.95,
  y_min = -Inf,
  y_max = Inf,
  ...
)

Arguments

object

A conformalRegressorByHorizon object

y_hat

Predicted values (one per horizon)

confidence

Confidence level(s) between 0 and 1 (e.g., 0.95 for 95 percent)

y_min

The minimum value to include in prediction intervals

y_max

The maximum value to include in prediction intervals

...

Ignored

Value

A data frame with lower and upper bounds for each confidence level

Author(s)

Resul Akay

Objects exported from other packages

Description

These objects are imported from other packages. Follow the links below to see their documentation.

forecast

autolayer, autoplot

generics

accuracy, forecast

magrittr

%<>%, %>%

Grouped sales data from an Australian Retailer

Description

A dataset containing 42 products' sales

Usage

retail

Format

A data class of "tbl_df", "tbl", "data.frame" with 13986 rows and 3 columns:

date

date

item

products

value

sales

Source

https://robjhyndman.com/data/ausretail.csv

Sales data from an Australian Retailer in time series format

Description

A dataset containing 42 products' sales

Usage

retail_wide

Format

An object of class mts (inherits from ts, matrix) with 333 rows and 43 columns.

This data set is the wide format of retail data.

Source

https://robjhyndman.com/data/ausretail.csv

Split a time series into training and testing sets

Description

Split a time series into training and testing sets

Usage

split_ts(y, test_size = 10)

Arguments

y

A univariate time series

test_size

The number of observations to keep in the test set

Value

A list with train and test elements

Author(s)

Resul Akay

Examples


dlist <- split_ts(retail_wide[,1], test_size = 12)

Suggested methods for ARml

Description

Suggested methods for ARml

Usage

suggested_methods()

Value

A character vector of Suggested methods

Author(s)

Resul Akay

Examples


suggested_methods()