Exploratory Data Analysis (EDA) is the initial and an important phase of data analysis. Through this phase, analysts/modelers will have a first look of the data, and thus generate relevant hypothesis and decide next steps. However, the EDA process could be a hassle at times. This R package aims to automate most of data handling and visualization, so that users could focus on studying the data and extracting insights.
The package can be installed directly from CRAN.
install.packages("DataExplorer")
However, the latest stable version (if any) could be found on GitHub, and installed using remotes package.
if (!require(remotes)) install.packages("remotes")
remotes::install_github("boxuancui/DataExplorer")
If you would like to install the latest development version, you may install the dev branch.
if (!require(remotes)) install.packages("remotes")
remotes::install_github("boxuancui/DataExplorer", ref = "develop")
The package is extremely easy to use. Almost everything could be done in one line of code. Please refer to the package manuals for more information. You may also find the package vignettes here.
To get a report for the airquality dataset:
library(DataExplorer)
create_report(airquality)
To get a report for the diamonds dataset from ggplot2 package:
library(DataExplorer)
library(ggplot2)
create_report(diamonds)
You may also run all the plotting functions individually for your analysis, e.g.,
library(DataExplorer)
library(ggplot2)
## View distribution of all discrete variables
plot_bar(diamonds)
## View distribution of cut only
plot_bar(diamonds$cut)
## View correlation of all discrete varaibles
plot_correlation(diamonds, type = "discrete")
## View distribution of all continuous variables
plot_histogram(diamonds)
## View distribution of carat only
plot_histogram(diamonds$carat)
## View correlation of all continuous varaibles
plot_correlation(diamonds, type = "continuous")
## View overall correlation heatmap
plot_correlation(diamonds)
## View distribution of missing values for airquality data
missing_data <- plot_missing(airquality) # missing data profile will be returned
missing_data
To visualize distributions based on another variable, you may do the following:
library(DataExplorer)
## View iris continuous distribution based on each Species
plot_boxplot(iris, "Species")
## View iris continuous distribution based on different buckets of Sepal.Length
plot_boxplot(iris, "Sepal.Length")
## Scatterplot Ozone against all other airquality features
# Set some features to factor
for (i in c("Month", "Day")) airquality[[i]] <- as.factor(airquality[[i]])
# Plot scatterplot
# Note: discrete and continuous charts are plotted on separate pages!
plot_scatterplot(airquality, "Ozone")
Sometimes, discrete variables are messy, e.g., too many imbalanced categories, extremely skewed categorical distribution. You may use group_category function to help you group the long tails.
library(DataExplorer)
library(ggplot2)
data(diamonds)
## View original distribution of variable clarity
diamonds <- data.table(diamonds)
table(diamonds$clarity)
## Trial and error without updating: Group bottom 20% clarity based on frequency
group_category(diamonds, "clarity", 0.2)
## Group bottom 30% clarity and update original dataset
group_category(diamonds, "clarity", 0.3, update = TRUE)
## View distribution after updating
table(diamonds$clarity)
## Group bottom 20% cut using value of carat
table(diamonds$cut)
group_category(diamonds, "cut", 0.2, measure = "carat", update = TRUE)
table(diamonds$cut)
Note: this function works with data.table objects only. If you are working with data.frame, please add data.table class to your object and then remove it later. See example below.
library(DataExplorer)
## Set data.frame object to data.table
USArrests <- data.table(USArrests)
## Collapse bottom 10% UrbanPop based on frequency
group_category(USArrests, "UrbanPop", 0.1, update = TRUE)
## Set object back to data.frame
class(USArrests) <- "data.frame"
plot_str: Plot data structure in network graph.drop_columns: Quickly drop variables with either column index or column names. (data.table only)set_missing: Quickly set all missing observations to a value. (data.table only)split_columns: Split data into two objects: discrete and continous.