Forward selection by rank

A model is trained and performance metric computed by including increasing numbers of features in the model. The features to be included in each step are defined by their rank, which is computed from another variable e.g. VIP score. An "optimal"subset of features is suggested by minimising the input performance metric.

Usage

forward_selection_by_rank(
  min_no_vars = 1,
  max_no_vars = 100,
  step_size = 1,
  factor_name,
  variable_rank,
  ...
)

Arguments

min_no_vars

(numeric) The minimum number of variables to include in the model. The default is 1.

max_no_vars

(numeric) The maximum number of variables to include in the model. The default is 100.

step_size

(numeric) The incremental change in number of features in the model. The default is 1.

factor_name

(character) The name of a sample-meta column to use.

variable_rank

(numeric, integer) The values used to rank the features.

...

Additional slots and values passed to struct_class.

Value

A forward_selection_by_rank object with the following output slots:

metric	(data.frame) The value of the computed metric for each model. For nested models the metric is averaged.
results	(data.frame) The predicted outputs from collated from all models computed during forward selection.
chosen_vars	(numeric, integer) The column number of the variables chosen for the best performing model.
smoothed	(numeric) The value of the performance metric for each evaluated model after smoothing.
searchlist	(numeric) The maxmimum rank of features included in each model.

Inheritance

A forward_selection_by_rank object inherits the following struct classes:

[forward_selection_by_rank] >> [resampler] >> [iterator] >> [struct_class]

Examples

M = forward_selection_by_rank(
      min_no_vars = 1,
      max_no_vars = 100,
      step_size = 1,
      factor_name = "V1",
      variable_rank = 1)

# some data
D = MTBLS79_DatasetExperiment(filtered=TRUE)

# normalise, impute and scale then remove QCs
P = pqn_norm(qc_label='QC',factor_name='Class') +
    knn_impute(neighbours=5) +
    glog_transform(qc_label='QC',factor_name='Class') +
    filter_smeta(mode='exclude',levels='QC',factor_name='Class')
P = model_apply(P,D)
D = predicted(P)

# forward selection using a PLSDA model
M = forward_selection_by_rank(factor_name='Class',
                             min_no_vars=2,
                             max_no_vars=11,
                             variable_rank=1:2063) *
    (mean_centre() + PLSDA(number_components=1,
                           factor_name='Class'))
M = run(M,D,balanced_accuracy())