R/data_viz.R

In Curious-Joe/mixBag: Collection of miscellaneous useful utility functions

# QUANTILE BASED TARGET LABEL RATIO ----

#' Ratio of target categories in binned numeric predictor
#'
#' @description Plot the ratio of categories of target labels inside different bins created based on quantiles for numeric features
#'
#' @param data Name of the data frame object
#' @param x_feat Name of the predictor of the interest
#' @param y_feat Name of the target feature
#' @param quantile Number of quantiles to be used to bin the numeric feature
#'
#' @return
#'
#' Returns three outputs:
#' 1. Quantile boundaries,
#' 2. Total records in each quantiles based bin,
#' 3. A line plot showing ratio in different bins. The output of the function can be passed to plotly to generate interactive plot
#'
#' @examples
#'
#' library(tidyverse)
#' qtile_plot(iris, Sepal.Length, Species, 4)
#'
#'
#' @export

qtile_plot = function(data, x_feat, y_feat, quantile = 4){

  x_str <- rlang::enquo(x_feat) %>% rlang::get_expr()
  y_str <- rlang::enquo(y_feat) %>% rlang::get_expr()

  Q = dplyr::pull(data, {{x_feat}}) %>%
    stats::quantile(probs = seq(0, 1, 1/quantile))

  if(length(unique(Q)) < quantile+1){
    stop(paste0(x_feat,
                " doesn't have enough unique values. Lower the number of quantiles."))
  }

  cat("Quantile boundaries:\n")
  print(Q)

  TILES <- cut(dplyr::pull(data, {{x_feat}}), breaks = c(Q), labels = paste0("Q", seq(1, quantile, 1)),
               include.lowest = T)

  cat("\nQuantile wise total counts:\n")
  print(table(TILES))

  data$TILES <- TILES

  data %>% dplyr::select(TILES, {{y_feat}}) %>% table()

  data %>%
    dplyr::count({{y_feat}}, TILES) %>%
    dplyr::group_by(TILES) %>%
    dplyr::mutate(PERCENT = n/sum(n)) %>%
    ggplot2::ggplot(ggplot2::aes(x = TILES, y = PERCENT, group = {{y_feat}}, color = {{y_feat}})) +
    ggplot2::geom_point() +
    ggplot2::geom_line()  +
    ggplot2::labs(title = paste0("Ratio of ", y_str, " Categories in ", quantile, " Quantile Groups of ", x_str)) +
    ggplot2::theme_minimal()

}

# BOX PLOTS FOR NUMERIC FEATURES SPLIT IN TARGET FEAT CATEGORIS ----

#' A function to display or save box plots created on all numeric features.
#'
#' @description Creates (or saves as png if asked) boxplots for all numeric features. Split the box plots based on target feature categories.
#'
#' @param dataset Name of the data frame object
#' @param classVar Name of the target feature
#' @param order A vector listing the target feature labels in the desired order. To use default order leave this parameter. Default value NULL.
#' @param colors A vector listing which color to use to represent which target feature label. To have the function pick color leave this parameter. Deafult value NULL.
#' @param loc A string with the directory where you want to save the plots. If no location is provided the plots will be created and displayed but not stored as image files.
#'
#' @return
#'
#' Returns three outputs:
#' 1. If loc = NULL, returns plots and displays in the RStudio Plots window,
#' 2. If loc has location provided, creates and saves the plots. Doesn't display in RStudio.
#'
#' @examples
#'
#' library(tidyverse)
#' boxPlot(dataset = iris,
#'      classVar = Species,
#'           order = c("virginica", "versicolor", "setosa"),
#'           colors = c("virginica" = "#32a897", "versicolor" = "#328bab", "setosa" = "#8031ad"))
#'
#'
#' @export

boxPlot <- function(dataset, classVar, order = NULL, colors = NULL, loc = NULL) {

  x <- rlang::enquo(classVar) %>% rlang::get_expr()
  nLevels <- dplyr::select(dataset, {{classVar}}) %>% unique() %>% nrow()
  # producing random color n case no color is provided
  if(is.null(colors)){
    colors = randomcoloR::randomColor(nLevels, luminosity = "bright")
  }

  if(is.null(loc)){
    for(i in names(dplyr::select(dataset, where(is.numeric)))) {
      print(dataset %>%
              ggplot2::ggplot(ggplot2::aes_string(x, {i}, color = x)) +
              ggplot2::geom_boxplot(show.legend = FALSE) +
              ggplot2::scale_x_discrete(limits = order) +
              ggplot2::scale_color_manual(values = colors) +
              ggplot2::labs(title = paste0("Distribution of ", i, " in Different ", x, " Categories")) +
              ggplot2::theme_minimal())
    }
  } else{
    for(i in names(dplyr::select(dataset, where(is.numeric)))) {
      grDevices::png(paste0(loc, "/boxplot_", i, ".PNG"), width = 627, height = 453)
      plot <- dataset %>%
        ggplot2::ggplot(ggplot2::aes_string(x, {i}, color = x)) +
        ggplot2::geom_boxplot(show.legend = FALSE) +
        ggplot2::scale_x_discrete(limits = order) +
        ggplot2::scale_color_manual(values = colors) +
        ggplot2::labs(title = paste0("Distribution of ", i, " in Different ", x, " Categories")) +
        ggplot2::theme_minimal()
      print(plot)
      grDevices::dev.off()
    }
  }
}

# BAR PLOTS FOR CATEGORICAL FEATURES SPLIT IN TARGET FEAT CATEGORIS ----


#' A util function to plot a customized ggplot
#' @noRd
barPlot <- function(data, x, y, fill, position, colors){
  data %>%
    ggplot2::ggplot(aes_string(x=x, y=y, fill=fill)) +
    ggplot2::geom_bar(position=position, stat="identity") +
    ggplot2::scale_fill_manual(values = colors) +
    ggplot2::labs(title = paste0(x, " in Different Categories of ", fill))
}

# BAR PLOTS FOR CATEGORICAL FEATURES SPLIT IN TARGET FEAT CATEGORIS ----

#' A function to display or save density plots created on all categorical features.
#'
#' @description Creates (or saves as png if asked) bar plots for all density features. Organize the bar plots based on target feature categories and desired barplot types of: dodge, stack, fill.
#'
#' @param dataset Name of the data frame object
#' @param classVar Name of the target feature
#' @param order A vector listing the target feature labels in the desired order. To use default order leave this parameter. Default value NULL.
#' @param colors A vector listing which color to use to represent which target feature label. To have the function pick color leave this parameter. Default value NULL.
#' @param barType A string value indicating bar plot type i.e. stack, dodge, fill.
#' @param loc A string with the directory where you want to save the plots. If no location is provided the plots will be created and displayed but not stored as image files.
#'
#' @return
#'
#' Returns three outputs:
#' 1. If loc = NULL, returns plots and displays in the RStudio Plots window,
#' 2. If loc has location provided, creates and saves the plots. Doesn't display in RStudio.
#'
#' @importFrom rlang enquo get_expr
#' @import dplyr ggplot2
#' @importFrom forcats fct_relevel
#' @importFrom randomcoloR randomColor
#' @importFrom grDevices dev.off png
#' @importFrom rlang :=
#'
#'
#' @examples
#' library(dplyr)
#' library(ggplot2)
#' biv_bar_plot(dataset = iris %>%
#' mutate(sepal_width_cat = ifelse(Sepal.Width < mean(iris$Sepal.Width), 'Low', 'High')),
#' classVar = Species)
#'
#' @export
#'

biv_bar_plot <- function(dataset, classVar, order = NULL,
                         colors = NULL, barType = "dodge", loc = NULL) {

  x <- rlang::enquo(classVar) %>% rlang::get_expr()
  nLevels <- dplyr::select(dataset, {{classVar}}) %>% unique() %>% nrow()

  # reordering levels in factor
  if(!is.null(order)){
    dataset <- dataset %>%
      dplyr::mutate({{classVar}} := forcats::fct_relevel({{classVar}}, order))
  }

  # producing random color n case no color is provided
  if(is.null(colors)){
    colors = randomcoloR::randomColor(nLevels, luminosity = "bright")
  }

  # fetching categorical feature names
  cols <- names(dplyr::select(dataset %>%
                                dplyr::select(-all_of(x)), !where(is.numeric)))

  {
    # fetching bar plot type
    if(!barType %in% c("fill", "stack", "dodge")){
      barType = readline(prompt = "Please enter one of these options fill, stack, dodge: ")
    }

    # print("plotting...")
    if(is.null(loc)){
      for(i in cols) {
        print(dataset %>%
                dplyr::select(x, all_of(i)) %>%
                table() %>%
                data.frame() %>%
                barPlot(x = i, y = "Freq", fill = x, position = barType, colors)
        )
      }
    } else{

      for(i in cols) {

        png(paste0(loc, "/barplot_", i, ".PNG"), width = 627, height = 453)
        plot <- dataset %>%
          dplyr::select(x, all_of(i)) %>%
          table() %>%
          data.frame() %>%
          barPlot(x = i, y = "Freq", fill = x, position = barType, colors)
        print(plot)
        dev.off()
      }
    }
  }
}

# DENSITY PLOTS FOR NUMERIC FEATURES SPLIT IN TARGET FEAT CATEGORIS ----

#' A function to display or save density plots created on all numeric features.
#'
#' @description Creates (or saves as png if asked) density plots for all numeric features. Splits the density plots based on target feature categories.
#'
#' @param dataset Name of the data frame object
#' @param classVar Name of the target feature
#' @param order A vector listing the target feature labels in the desired order. To use default order leave this parameter. Default value NULL.
#' @param colors A vector listing which color to use to represent which target feature label. To have the function pick color leave this parameter. Deafult value NULL.
#' @param loc A string with the directory where you want to save the plots. If no location is provided the plots will be created and displayed but not stored as image files.
#'
#' @return
#'
#' Returns three outputs:
#' 1. If loc = NULL, returns plots and displays in the RStudio Plots window,
#' 2. If loc has location provided, creates and saves the plots. Doesn't display in RStudio.
#'
#' @examples
#'
#' library(tidyverse)
#' densePlot(dataset = iris,
#' classVar = Species,
#' order = c("virginica", "versicolor", "setosa"),
#' colors = c("virginica" = "#32a897", "versicolor" = "#328bab", "setosa" = "#8031ad"),
#' loc = NULL)
#'
#'
#' @export

densePlot <- function(dataset, classVar, order = NULL, colors = NULL, loc = NULL) {

  x <- rlang::enquo(classVar) %>% rlang::get_expr()
  nLevels <- dplyr::select(dataset, {{classVar}}) %>% unique() %>% nrow()

  # producing random color n case no color is provided
  if(is.null(colors)){
    colors = randomcoloR::randomColor(nLevels, luminosity = "bright")
  }

  print("plotting...")
  if(is.null(loc)){
    for(i in names(dplyr::select(dataset, where(is.numeric)))) {
      print(dataset %>%
              ggplot2::ggplot(ggplot2::aes_string(i, fill = x, color = x)) +
              ggplot2::geom_density(alpha = 0.3, na.rm = TRUE) +
              ggplot2::scale_fill_manual(values = colors) +
              ggplot2::labs(title = paste0("Distribution of ", i, " Based on ", x)) +
              ggplot2::theme_minimal())
    }
  } else{

    for(i in names(dplyr::select(dataset, where(is.numeric)))) {

      grDevices::png(paste0(loc, "/denseplot_", i, ".PNG"), width = 627, height = 453)
      plot <- dataset %>%
        ggplot2::ggplot(ggplot2::aes_string({i}, fill = x, color = x)) +
        ggplot2::geom_density(alpha = 0.3, na.rm = TRUE) +
        ggplot2::scale_fill_manual(values = colors) +
        ggplot2::labs(title = paste0("Distribution of ", i, " Based on ", x)) +
        ggplot2::theme_minimal()
      print(plot)
      grDevices::dev.off()
    }
  }
}