auto_theta: Auto Tehta

In yvesmauron/univariate-time-series-forecasting: Intelligent Algorithms to be Integrated in Business Intelligence solutions.

Description

Combination of univariate theta methods, using a combination operator of choice for the mean forecasts and prediction intervals.

Usage

auto_theta(
  y,
  h,
  level = 95,
  point_combination = "weighted_average",
  pi_combination_upper = "median",
  pi_combination_lower = "median",
  pool_limit = 3,
  error_fun = "rmse",
  weight_fun = "inverse",
  val_h = h,
  sov_only = F,
  max_years = 30,
  val_min_years = 4,
  cv_min_years = 5,
  cv_max_samples = 3,
  allow_negatives = F,
  ...
)

Arguments

y	Time series object with historical values.
h	Horizons to be predicted.
point_combination	Point forecast combination operator. Optional, default meidan.
pi_combination_upper	combination operator of the upper bound of the prediction intervals. Optional, default max.
pi_combination_lower	combination operator for the lower bounds of the prediction intervals. Optional, default min.
pool_limit	number of methods selected from the pool to reach the final forecasts. Optional, default length(methods).
error_fun	error function to determine the validation performance Optional, default rmse.
weight_fun	weight function for calculating the definitive weights for combination. Optional, default inverse
val_h	The horizon for the validation samples. Optional, default h.
sov_only	Flag indicating whether only single origin validation should be considered. Optional, default TRUE.
max_years	Maxmium of years to consider during model fitting. Optional, default 30.
val_min_years	Minimum years required to conduct single origin validation. Optional, default 4.
cv_min_years	Minimum years required to conduct cross-origin validation. Optional, default 5.
cv_max_samples	Maximum samples that should be considered during cross-validation, i.e. 3 indicates the algorithm validates from 3 origins. Optional, default 3.
allow_negatives	Flag indicating whether to allow negative values or not. Optional, default FALSE.
...	passed to the forecasting functions
levels	Prediction interval levels. Optional, default c(95).
methods	Methods to be combined. Optional, default auto_ets, auto_arima, auto_dotm.

Value

combined forecasts of time series y including confidence intervals

yvesmauron/univariate-time-series-forecasting documentation built on March 2, 2020, 12:20 a.m.