select_model_hmnc: Select the optimal Harmonic Regression Model
In placeboo/amplify: Forecasting Daily Time Seriers with Double Seasonal Patterns

Description Usage Arguments Value Examples

double seasonal harmonic regression is implemented. If grid.search = TRUE, grid search is applied by searching the number of Fourier terms for yearly seasonality. search.length and length.out are for grid search. The search is based on the minimum mean absolute percentage error with parallel computing. More details about parallel computing can be found in doParallel.

select_model_hmnc(
  train.y,
  valid.y = NULL,
  s1 = 7,
  s2 = 365.25,
  grid.search = FALSE,
  search.length = c(20, 40),
  length.out = 5
)

`train.y`	A numeric vector for training.
`valid.y`	A numeric vector for validating If `grid.search = FALSE`, `valid.y = NULL`.
`s1`	Period of the shorter seasonal period.
`s2`	Period of the longer seasonal period.
`grid.search`	If TRUE, a grid search is applied.
`search.length`	Grid search parameter. Only used if `grid.search = TRUE`. It is the range of the number of the Fourier term of yearly seasonality.
`length.out`	Grid search parameter. Only used if `grid.search = TRUE`. It is the desired length of search sequence.

A list contains:

model. An object of class forecast. The model is built by combining training and testing;
K. A vector display the number of Fourier term for weekly seasonality and yearly seasonality.
cv. A data frame of searched parameters with respect to its distance measures (see measure_dist)

data(tickets)
data.ls = train_test_split(tickets, var = "date", train.window = c(20140701, 20180630), test.window = c(20180701, 20190630))
train.y = data.ls$train.dat[,2]
valid.y = data.ls$test.dat[,2]
select.ls = select_model_hmnc(train.y, valid.y, grid.search = TRUE, length.out = 5)