RemixAutoML (version 0.5.4)

CategoricalEncoding: CategoricalEncoding

Description

Categorical encoding for factor and character columns

Usage

CategoricalEncoding(
  data = NULL,
  ML_Type = "classification",
  GroupVariables = NULL,
  TargetVariable = NULL,
  Method = NULL,
  SavePath = NULL,
  Scoring = FALSE,
  ImputeValueScoring = NULL,
  ReturnFactorLevelList = TRUE,
  SupplyFactorLevelList = NULL,
  KeepOriginalFactors = TRUE
)

Arguments

data

Source data

ML_Type

Only use with Method "credibility'. Select from 'classification' or 'regression'.

GroupVariables

Columns to encode

Method

Method to utilize. Choose from 'm_estimator', 'credibility', 'woe', 'target_encoding', 'poly_encode', 'backward_difference', 'helmert'

SavePath

Path to save artifacts for recreating in scoring environments

Scoring

Set to TRUE for scoring mode.

ImputeValueScoring

If levels cannot be matched on scoring data you can supply a value to impute the NA's. Otherwise, leave NULL and manage them outside the function

ReturnFactorLevelList

TRUE by default. Returns a list of the factor variable and transformations needed for regenerating them in a scoring environment. Alternatively, if you save them to file, they can be called for use in a scoring environment.

SupplyFactorLevelList

The FactorCompenents list that gets returned. Supply this to recreate features in scoring environment

KeepOriginalFactors

Defaults to TRUE. Set to FALSE to remove the original factor columns

TargetVariabl

Target column name

See Also

Other Feature Engineering: AutoDataPartition(), AutoDiffLagN(), AutoHierarchicalFourier(), AutoInteraction(), AutoLagRollStatsScoring(), AutoLagRollStats(), AutoTransformationCreate(), AutoTransformationScore(), AutoWord2VecModeler(), AutoWord2VecScoring(), CreateCalendarVariables(), CreateHolidayVariables(), DummifyDT(), H2OAutoencoderScoring(), H2OAutoencoder(), ModelDataPrep(), TimeSeriesFill()

Examples

# NOT RUN {
# Create fake data with 10 categorical
data <- RemixAutoML::FakeDataGenerator(
  Correlation = 0.85,
  N = 1000000,
  ID = 2L,
  ZIP = 0,
  FactorCount = 10L,
  AddDate = FALSE,
  Classification = TRUE,
  MultiClass = FALSE)

# Take your pick
Meth <- c('m_estimator',
          'credibility',
          'woe',
          'target_encoding',
          'poly_encode',
          'backward_difference',
          'helmert')

# Pass to function
MethNum <- 1

# Mock test data with same factor levels
test <- data.table::copy(data)

# Run in Train Mode
data <- RemixAutoML::CategoricalEncoding(
  data = data,
  ML_Type = "classification",
  GroupVariables = paste0("Factor_", 1:10),
  TargetVariable = "Adrian",
  Method = Meth[MethNum],
  SavePath = getwd(),
  Scoring = FALSE,
  ReturnFactorLevelList = FALSE,
  SupplyFactorLevelList = NULL,
  KeepOriginalFactors = FALSE)

# View results
print(data)

# Run in Score Mode by pulling in the csv's
test <- RemixAutoML::CategoricalEncoding(
  data = data,
  ML_Type = "classification",
  GroupVariables = paste0("Factor_", 1:10),
  TargetVariable = "Adrian",
  Method = Meth[MethNum],
  SavePath = getwd(),
  Scoring = TRUE,
  ImputeValueScoring = 222,
  ReturnFactorLevelList = FALSE,
  SupplyFactorLevelList = NULL,
  KeepOriginalFactors = FALSE)
# }
# NOT RUN {
# }