RemixAutoML (version 0.5.4)

AutoInteraction: AutoInteraction

Description

AutoInteraction creates interaction variables from your numerical features in your data. Supply a set of column names to utilize and set the interaction level. Supply a character vector of columns to exclude and the function will ignore those features.

Usage

AutoInteraction(
  data = NULL,
  NumericVars = NULL,
  InteractionDepth = 2,
  Center = TRUE,
  Scale = TRUE,
  SkipCols = NULL,
  Scoring = FALSE,
  File = NULL
)

Arguments

data

Source data.table

InteractionDepth

The max K in N choose K. If NULL, K will loop through 1 to length(NumVars). Default is 2 for pairwise interactions

Center

TRUE to center the data

Scale

TRUE to scale the data

SkipCols

Use this to exclude features from being created. An example could be, you build a model with all variables and then use the varaible importance list to determine which features aren't necessary and pass that set of features into this argument as a character vector.

Scoring

Defaults to FALSE. Set to TRUE for generating these columns in a model scoring setting

File

When Scoring is set to TRUE you have to supply either the .Rdata list with lookup values for recreating features or a pathfile to the .Rdata file with the lookup values. If you didn't center or scale the data then this argument can be ignored.

NumVars

Names of numeric columns (if NULL, all numeric and integer columns will be used)

See Also

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

Examples

# NOT RUN {
#########################################
# Feature Engineering for Model Training
#########################################

# Create fake data
data <- RemixAutoML::FakeDataGenerator(
  Correlation = 0.70,
  N = 50000,
  ID = 2L,
  FactorCount = 2L,
  AddDate = TRUE,
  ZIP = 0L,
  TimeSeries = FALSE,
  ChainLadderData = FALSE,
  Classification = FALSE,
  MultiClass = FALSE)

# Print number of columns
print(ncol(data))

# Store names of numeric and integer cols
Cols <-names(data)[c(which(unlist(lapply(data, is.numeric))),
                     which(unlist(lapply(data, is.integer))))]

# Model Training Feature Engineering
system.time(data <- RemixAutoML::AutoInteraction(
  data = data,
  NumericVars = Cols,
  InteractionDepth = 4,
  Center = TRUE,
  Scale = TRUE,
  SkipCols = NULL,
  Scoring = FALSE,
  File = getwd()))

# user  system elapsed
# 0.30    0.11    0.41

# Print number of columns
print(ncol(data))

########################################
# Feature Engineering for Model Scoring
########################################

# Create fake data
data <- RemixAutoML::FakeDataGenerator(
  Correlation = 0.70,
  N = 1000,
  ID = 2L,
  FactorCount = 2L,
  AddDate = TRUE,
  ZIP = 0L,
  TimeSeries = FALSE,
  ChainLadderData = FALSE,
  Classification = FALSE,
  MultiClass = FALSE)

# Print number of columns
print(ncol(data))

# Reduce to single row to mock a scoring scenario
data <- data[1L]

# Model Scoring Feature Engineering
system.time(data <- RemixAutoML::AutoInteraction(
  data = data,
  NumericVars = names(data)[
    c(which(unlist(lapply(data, is.numeric))),
      which(unlist(lapply(data, is.integer))))],
  InteractionDepth = 4,
  Center = TRUE,
  Scale = TRUE,
  SkipCols = NULL,
  Scoring = TRUE,
  File = file.path(getwd(), "Standardize.Rdata")))

# user  system elapsed
# 0.19    0.00    0.19

# Print number of columns
print(ncol(data))
# }