forestRK (version 0.0-5)

x.organizer: Numericizing a data frame of covariates from the original dataset via Binary or Numeric Encoding

Description

Takes the original data frame of covariates as an input (which may or may not be numeric), and converts it into a numericized data frame by applying either Binary or Numeric Encoding.

Binary Encoding for categorical features are recommended for tree ensembles when the cardinality of categorical feature is >= 1000; Numeric Encoding for categorical features are recommended for tree ensembles when the cardinality of categorical features is < 1000.

For more information about the Binary and Numeric Encoding and their effectiveness under different cardinality, please visit: https://medium.com/data-design/ visiting-categorical-features-and-encoding-in-decision-trees-53400fa65931

NOTE: In order to use other functions within the forestRK package, you must ensure that the numericized data frame of covariates (the x.organizer object) contains no missing record, that is, you have to remove any record containing NA or NaN prior to applying the x.organizer function.

Following is the summary of the data cleaning process with x.organizer():

1. remove all NA or NaN's from the data in hand. 2. split the data into a data frame that contains covariates of ALL data points, (BOTH training and test observations), and a vector that contains class types of the training observations; 3. apply the x.organizer to the big data frame of covariates of all observations. 4. split the x.organizer output into a training and a test set, as needed.

PROPER DATA CLEANING IS ABSOLUTELY NECESSARY FOR forestRK FUNCTIONS TO WORK!

Usage

x.organizer(x.dat = data.frame(), encoding = c("num","bin"))

Arguments

x.dat

a data frame storing covariates of each observation (can be either numeric or non-numeric) from the original data; x.dat should contain no NA or NaN. The rownames of x.dat should be numerical index for each observations.

encoding

type of encoding done for the categorical features; "num" stands for Numeric Encoding, and "bin" stands for Binary Encoding. When the data in question only has numeric features, then the user can select either one of "num" or "bin", and the x.organizer function will just return the original numeric dataset.

Value

A numericized data frame of the covariates from the original data obtained via either Numeric or Binary Encoding.

See Also

y.organizer

Examples

# NOT RUN {
  ## example: iris dataset
  library(forestRK) # load the package forestRK

  ## Basic Procedures
  ## 1. Apply x.organizer to a data frame that stores covariates of
  ## ALL observations (BOTH training and test observations)
  ## 2. Split the output from 1 into a training and a test set, as needed

  # note: iris[,1:4] are the columns of the iris dataset that stores
  # covariate values

  # covariates of training data set
  x.train <- x.organizer(iris[,1:4], encoding = "num")[c(1:25,51:75,101:125),]
# }