Wrapper function to compute the benchmark

regional_benchmark() is a wrapper function calling a number of functions. See details.

regional_benchmark(
  regions = c("ALLSAC", "SFE", "K", "NC", "NCC", "SCC", "SC", "SJT"),
  LRN_IDS,
  TUNELENGTH,
  INNER,
  ITERS,
  PROB,
  NU,
  REPS,
  PREPROC,
  FINAL,
  PATH,
  REDUCED,
  MES,
  INFO,
  FS,
  FS_NUM
)

Arguments

regions	character, vector of region identifiers (e.g. "SFE")
LRN_IDS	character, vector of learner identifiers (e.g. "classif.randomForest")
TUNELENGTH	numeric, controls the number of hyper-parameters for discrete tuning
INNER	ResampleDesc, the inner folds for the nested resampling
ITERS	numeric, controls the number of tries for the random tuning
PROB	logical, selects the type of output, if TRUE probabilities, if FALSE response
NU	numeric, number of folds for the \(\nu\)-fold cross-validation
REPS	numeric, number of repetitions for the repeated \(\nu\)-fold cross-validation
PREPROC	character, vector of preprocessing identifiers (e.g. "scale")
FINAL	logical, selects the type of runs, if FALSE trains, else predict
PATH	character or file.path, path to the output directory
REDUCED	logical, legacy option
MES	list, list of measures from the mlr package, the first one is optimized against
INFO	logical, controls the information printed by the training process
FS	logical, if TRUE activates the feature selection
FS_NUM	numeric, number of features to select if FS is TRUE

Value

a list of mlr benchmark results

Details

Here is a some pseudo-code that explains what is happening behind the scenes.

1. Skip. Because regional_benchmark() is called inside the for-loop for (FS_NUM in FS_NUM_LIST) (see above), if FINAL is #' TRUE, regional_benchmark() skips the region it does not need to calculate the final models.

2. Data loading. This handled by get_training_data()

3. Data formatting. This is handled by fmt_labels(), sanitize_data() and get_coords().

4. Feature selection. If FINAL is TRUE the selected features are retrieved from get_bestBMR_tuning_results(). If FINAL is FALSE the selected features are derived from transformed training data using get_ppc() and preproc_data(). The resulting transformed data are filtered for correlation higher than 0.95 with caret::findCorrelation(). Then, 500 subsampled mlr Tasks are created with mlr::makeResampleDesc(), mlr::makeClassifTask(), mlr::makeResampleInstance() and mlr::filterFeatures(). The #' FS_NUM most commonly select features across the 500 realizations are selected.

5. Pre-processing. The target and training data are transformed using get_ppc() on the target data and preproc_data() on the #' training data. SMOTE is applied using get_smote_data() and get_smote_coords() which both call resolve_class_imbalance().

6. Tasks. Tasks are obtained using mlr::makeClassifTask().

7. Learners. Learners are constructed using get_learners() or get_final_learners().

8. Compute benchmark. The benchmark is run with compute_final_model() or compute_benchmark() (which needs to retrieve the outer #' folds of the nested resampling with get_outers()).