Constructor for the dorem control object

dorem_control(
  weights = NULL,
  na.rm = TRUE,
  loss_func = dorem_loss_func,
  link_func = dorem_no_link_func,
  perf_func = dorem_perf_func,
  optim_method = valid_optimization_methods(),
  optim_maxit = 1000,
  optim_VTR = -Inf,
  optim_grid_n = 5,
  optim_trace = FALSE,
  coefs_start = NULL,
  coefs_lower = NULL,
  coefs_upper = NULL,
  cv_repeats = NULL,
  cv_folds = NULL,
  shuffle = FALSE,
  iter = TRUE,
  seed = NULL
)

Arguments

weights

Default is NULL

na.rm

Default is TRUE

loss_func

Function used in optimization objective function. Default is dorem_loss_func which returns MSE

link_func

Function used to convert predictions. Default is dorem_no_link_func which returns the original value

perf_func

Function used to quantify model fit. Default is dorem_perf_func

optim_method

Optimization method. The list of supported methods is c("L-BFGS-B", "DE", "CMA-ES", "gridSearch")

optim_maxit

Maximal number of iterations for the optimization method. Default is 1000

optim_VTR

Value To Reach in the optimization method. Default is -Inf

optim_grid_n

Number of search values for each predictor used in gridSearch method. Default is 5

optim_trace

Should optimization trace be shown? Default is FALSE

coefs_start

Starting values for coefficients

coefs_lower

Lower bound for coefficients

coefs_upper

Upper bound for coefficients

cv_repeats

Number of CV repeats

cv_folds

Number of CV folds

shuffle

Should shuffle be performed? Default is FALSE

iter

Should iter be shown? Default is TRUE

seed

Random number seed

Examples

data("bike_score") banister_model <- dorem( Test_5min_Power ~ BikeScore, bike_score, method = "banister", control = dorem_control( cv_folds = 3, cv_repeats = 5, shuffle = TRUE ) )
#> Performing banister method using L-BFGS-B optimization
#> Training the model...
#> Cross-validating the model using 5 repeats of 3 folds
#> Fold1.Rep1...
#> Fold2.Rep1...
#> Fold3.Rep1...
#> Fold1.Rep2...
#> Fold2.Rep2...
#> Fold3.Rep2...
#> Fold1.Rep3...
#> Fold2.Rep3...
#> Fold3.Rep3...
#> Fold1.Rep4...
#> Fold2.Rep4...
#> Fold3.Rep4...
#> Fold1.Rep5...
#> Fold2.Rep5...
#> Fold3.Rep5...
#> Training the model using shuffled predictors...
#> Done!