Data Transformation

knitr::opts_chunk$set(collapse = TRUE, comment = "", out.width = "600px", dpi = 70)
options(tibble.print_min = 4L, tibble.print_max = 4L)

library(dlookr)
library(dplyr)
library(ggplot2)

Preface

After you have acquired the data, you should do the following:

The dlookr package makes these steps fast and easy:

This document introduces data transformation methods provided by the dlookr package. You will learn how to transform of tbl_df data that inherits from data.frame and data.frame with functions provided by dlookr.

dlookr increases synergy with dplyr. Particularly in data transformation and data wrangle, it increases the efficiency of the tidyverse package group.

datasets

To illustrate the basic use of data transformation in the dlookr package, I use a Carseats dataset. Carseats in the ISLR package is simulation dataset that sells children's car seats at 400 stores. This data is a data.frame created for the purpose of predicting sales volume.

library(ISLR)
str(Carseats)

The contents of individual variables are as follows. (Refer to ISLR::Carseats Man page)

When data analysis is performed, data containing missing values is often encountered. However, Carseats is complete data without missing. Therefore, the missing values are generated as follows. And I created a data.frame object named carseats.

carseats <- ISLR::Carseats

suppressWarnings(RNGversion("3.5.0"))
set.seed(123)
carseats[sample(seq(NROW(carseats)), 20), "Income"] <- NA

suppressWarnings(RNGversion("3.5.0"))
set.seed(456)
carseats[sample(seq(NROW(carseats)), 10), "Urban"] <- NA

Data Transformation

dlookr imputes missing values and outliers and resolves skewed data. It also provides the ability to bin continuous variables as categorical variables.

Here is a list of the data conversion functions and functions provided by dlookr:

Imputation of missing values

imputes the missing value with imputate_na()

imputate_na() imputes the missing value contained in the variable. The predictor with missing values support both numeric and categorical variables, and supports the following method.

In the following example, imputate_na() imputes the missing value of Income, a numeric variable of carseats, using the "rpart" method. summary() summarizes missing value imputation information, and plot() visualizes missing information.

if (requireNamespace("rpart", quietly = TRUE)) {
  income <- imputate_na(carseats, Income, US, method = "rpart")

  # result of imputation
  income

  # summary of imputation
  summary(income)

  # viz of imputation
  plot(income)
} else {
  cat("If you want to use this feature, you need to install the rpart package.\n")
}

The following imputes the categorical variable urban by the "mice" method.

library(mice)

urban <- imputate_na(carseats, Urban, US, method = "mice")

# result of imputation
urban

# summary of imputation
summary(urban)

# viz of imputation
plot(urban)

Collaboration with dplyr

The following example imputes the missing value of the Income variable, and then calculates the arithmetic mean for each level of US. In this case, dplyr is used, and it is easily interpreted logically using pipes.

# The mean before and after the imputation of the Income variable
carseats %>%
  mutate(Income_imp = imputate_na(carseats, Income, US, method = "knn")) %>%
  group_by(US) %>%
  summarise(orig = mean(Income, na.rm = TRUE),
            imputation = mean(Income_imp))

Imputation of outliers

imputes thr outliers with imputate_outlier()

imputate_outlier() imputes the outliers value. The predictor with outliers supports only numeric variables and supports the following methods.

imputate_outlier() imputes the outliers with the numeric variable Price as the "capping" method, as follows. summary() summarizes outliers imputation information, and plot() visualizes imputation information.

price <- imputate_outlier(carseats, Price, method = "capping")

# result of imputation
price

# summary of imputation
summary(price)

# viz of imputation
plot(price)

Collaboration with dplyr

The following example imputes the outliers of the Price variable, and then calculates the arithmetic mean for each level of US. In this case, dplyr is used, and it is easily interpreted logically using pipes.

# The mean before and after the imputation of the Price variable
carseats %>%
  mutate(Price_imp = imputate_outlier(carseats, Price, method = "capping")) %>%
  group_by(US) %>%
  summarise(orig = mean(Price, na.rm = TRUE),
    imputation = mean(Price_imp, na.rm = TRUE))

Standardization and Resolving Skewness

Introduction to the use of transform()

transform() performs data transformation. Only numeric variables are supported, and the following methods are provided.

Standardization with transform()

Use the methods "zscore" and "minmax" to perform standardization.

carseats %>% 
  mutate(Income_minmax = transform(carseats$Income, method = "minmax"),
    Sales_minmax = transform(carseats$Sales, method = "minmax")) %>% 
  select(Income_minmax, Sales_minmax) %>% 
  boxplot()

Resolving Skewness data with transform()

find_skewness() searches for variables with skewed data. This function finds data skewed by search conditions and calculates skewness.

# find index of skewed variables
find_skewness(carseats)

# find names of skewed variables
find_skewness(carseats, index = FALSE)

# compute the skewness
find_skewness(carseats, value = TRUE)

# compute the skewness & filtering with threshold
find_skewness(carseats, value = TRUE, thres = 0.1)

The skewness of Advertising is 0.637. This means that the distribution of data is somewhat inclined to the left. So, for normal distribution, use transform() to convert to "log" method as follows. summary() summarizes transformation information, and plot() visualizes transformation information.

Advertising_log = transform(carseats$Advertising, method = "log")

# result of transformation
head(Advertising_log)
# summary of transformation
summary(Advertising_log)
# viz of transformation
plot(Advertising_log)

It seems that the raw data contains 0, as there is a -Inf in the log converted value. So this time, convert it to "log+1".

Advertising_log <- transform(carseats$Advertising, method = "log+1")

# result of transformation
head(Advertising_log)
# summary of transformation
summary(Advertising_log)
# viz of transformation
# plot(Advertising_log)

Binning

Binning of individual variables using binning()

binning() transforms a numeric variable into a categorical variable by binning it. The following types of binning are supported.

Here are some examples of how to bin Income using binning().:

# Binning the carat variable. default type argument is "quantile"
bin <- binning(carseats$Income)
# Print bins class object
bin
# Summarize bins class object
summary(bin)
# Plot bins class object
plot(bin)
# Using labels argument
bin <- binning(carseats$Income, nbins = 4,
              labels = c("LQ1", "UQ1", "LQ3", "UQ3"))
bin
# Using another type argument
binning(carseats$Income, nbins = 5, type = "equal")
binning(carseats$Income, nbins = 5, type = "pretty")

if (requireNamespace("classInt", quietly = TRUE)) {
  binning(carseats$Income, nbins = 5, type = "kmeans")
  binning(carseats$Income, nbins = 5, type = "bclust")
} else {
  cat("If you want to use this feature, you need to install the classInt package.\n")
}

# Extract the binned results
extract(bin)

# -------------------------
# Using pipes & dplyr
# -------------------------
library(dplyr)

carseats %>%
 mutate(Income_bin = binning(carseats$Income) %>% 
                     extract()) %>%
 group_by(ShelveLoc, Income_bin) %>%
 summarise(freq = n()) %>%
 arrange(desc(freq)) %>%
 head(10)

Optimal Binning with binning_by()

binning_by() transforms a numeric variable into a categorical variable by optimal binning. This method is often used when developing a scorecard model.

The following binning_by() example optimally binning Advertising considering the target variable US with a binary class.

library(dplyr)

# optimal binning using character
bin <- binning_by(carseats, "US", "Advertising")

# optimal binning using name
bin <- binning_by(carseats, US, Advertising)
bin

# summary optimal_bins class
summary(bin)

# performance table
attr(bin, "performance")

# visualize optimal_bins class
plot(bin)

# extract binned results
extract(bin) %>% 
  head(20)

Automated report

dlookr provides two automated data transformation reports:

Create a dynamic report using transformation_web_report()

transformation_web_report() create dynamic report for object inherited from data.frame(tbl_df, tbl, etc) or data.frame.

Contents of dynamic web report

The contents of the report are as follows.:

Some arguments for dynamic web report

transformation_web_report() generates various reports with the following arguments.

The following script creates a data transformation report for the tbl_df class object, heartfailure.

heartfailure %>%
  transformation_web_report(target = "death_event", subtitle = "heartfailure",
                            output_dir = "./", output_file = "transformation.html", 
                            theme = "blue")

Screenshot of dynamic report

knitr::include_graphics('img/transformation_web_title.jpg')

Create a static report using transformation_paged_report()

transformation_paged_report() create static report for object inherited from data.frame(tbl_df, tbl, etc) or data.frame.

Contents of static paged report

The contents of the report are as follows.:

Some arguments for static paged report

transformation_paged_report() generates various reports with the following arguments.

The following script creates a data transformation report for the data.frame class object, heartfailure.

heartfailure %>%
  transformation_paged_report(target = "death_event", subtitle = "heartfailure",
                              output_dir = "./", output_file = "transformation.pdf", 
                              theme = "blue")

Screenshot of static report

knitr::include_graphics('img/transformation_paged_cover.jpg')
knitr::include_graphics('img/transformation_paged_content.jpg')


Try the dlookr package in your browser

Any scripts or data that you put into this service are public.

dlookr documentation built on July 30, 2021, 9:07 a.m.