R/df_get_feature_bins.R

In BrandonRCopeland/DataScience: Provides Helper Functions for Data Science Analysis

#' Create bins for numeric features
#'
#' This function takes a matrix and returns a matrix with the designated number
#' of bins for the features.  Depending on the range and distribution of the numeric feature,
#' this function may return less than the designated number of bins.
#'
#' NOTE: This function currently only supports NUMERIC and/or CHARACTER data types. If you have
#'       other types of data, please filter them out before passing to the function.

#' @param data_ A data.frame containing features that need to be binned.
#' @param features_ A vector of the numeric feature names to be binned.
#' @param dataType "numeric" or "categorical"
#' @return A data.frame containing the feature name, bin number, min value, and max value
#' @export
df_get_feature_bins <- function(data_, features_, dataType = "numeric") {

  if(dataType == "numeric"){

    df.numeric.temp_ <- data_ %>%
      dplyr::select(dplyr::one_of(features_)) %>%
      dplyr::select_if(is.numeric) %>%
      tidyr::gather('feature', 'value') %>%
      dplyr::group_by(feature) %>%
      dplyr::mutate(bin = dplyr::ntile(value, 10L)) %>%
      dplyr::group_by(feature, bin) %>%
      dplyr::summarise(min = base::min(value),
                       max = base::max(value)) %>%
      dplyr::filter(dplyr::row_number() == 1 | min != max) %>% #edge cases where all bins are 0 min and max
      dplyr::mutate(min = base::ifelse(bin == base::min(bin, na.rm = TRUE), -Inf, min),
                    max = as.numeric(base::ifelse(bin == base::max(bin, na.rm = TRUE), Inf, max))) %>%
      dplyr::arrange(feature, bin) %>%
      dplyr::mutate(bin = as.character(dplyr::row_number()),
                    DataType = "numeric")

    return(df.numeric.temp_)
  }

  if(dataType == "categorical"){
    df.categorical.temp_ <- data_ %>%
      dplyr::select(dplyr::one_of(features_)) %>%
      dplyr::select_if(function(col) is.character(col) | is.factor(col)) %>%
      dplyr::mutate_all(funs(as.character)) %>%
      tidy::gather('feature', 'bin') %>%
      dplyr::distinct(feature, bin) %>%
      dplyr::mutate(min = NaN,
                    max = NaN,
                    DataType = "categorical") %>%
      dplyr::arrange(feature, bin)

    return(df.categorical.temp_)
  }

}