Pandas Profiling in Python

Pandas is a very vast library that offers many functions with the help of which we can understand our data. Pandas profiling provides a solution to this by generating comprehensive reports for datasets that have numerous features. These reports can be customized according to specific requirements. In this article, we will dive into this library's functionalities and explore its various features like:

• Installation of Pandas Profiling
• Importing Pandas Profiling
• Generating Profile Report
• Exploring Profile Report Generated
  • Overview
  • Variables
  • Correlations
  • Missing Values
  • Sample
• Saving the Profile Report

Installation of Pandas Profiling

Pandas Profiling can be easily installed using the following command

pip install pandas-profiling

The pandas_profiling library in Python includes a method named as ProfileReport() which generates a basic report on the input DataFrame. 

The report consists of the following:

• DataFrame overview,
• Each attribute on which DataFrame is defined,
• Correlations between attributes (Pearson Correlation and Spearman Correlation), and
• A sample of DataFrame.

Syntax :

pandas_profiling.ProfileReport(df, **kwargs)

Now, let's take an example, we will create our own data frame and will have a look at how pandas profiling can help in understanding the dataset more. Before that let us import the pandas_profiling.

Importing Pandas Profiling

# importing packages
import pandas as pd
from pandas_profiling import ProfileReport

Now, create the data frame or import any dataset.

# creating the dataframe
# dictionary of data
dct = {'ID': {0: 23, 1: 43, 2: 12, 3: 13, 
              4: 67, 5: 89, 6: 90, 7: 56, 
              8: 34}, 
       'Name': {0: 'Ram', 1: 'Deep', 2: 'Yash',
                3: 'Aman', 4: 'Arjun', 5: 'Aditya',
                6: 'Divya', 7: 'Chalsea',
                8: 'Akash' }, 
       'Marks': {0: 89, 1: 97, 2: 45, 3: 78,
                 4: 56, 5: 76, 6: 100, 7: 87,
                 8: 81}, 
       'Grade': {0: 'B', 1: 'A', 2: 'F', 3: 'C',
                 4: 'E', 5: 'C', 6: 'A', 7: 'B',
                 8: 'B'}
      }

# forming dataframe and printing
data = pd.DataFrame(dct)
print(data)

Output:

   ID     Name  Marks Grade
0  23      Ram     89     B
1  43     Deep     97     A
2  12     Yash     45     F
3  13     Aman     78     C
4  67    Arjun     56     E
5  89   Aditya     76     C
6  90    Divya    100     A
7  56  Chalsea     87     B
8  34    Akash     81     B

Generating Profile Report

For generating the profile report we will simply use the Profile Report from pandas_profile and input will the dataframe.

# forming ProfileReport and save
# as output.html file
profile = ProfileReport(data)
profile

Exploring the Profile Report Generated

Overview

Overview consists of 3 tabs, these are Overview, Alerts, and Reproduction.
The Overview consists of dataset statistics and variable types. Dataset statistics gives us information on number of variables, duplicates and missing values.

Next is the Alerts tab, which gives us information on the correlated variables. Also, about the unique values. Here, the data is small but if the dataset will be large then it will also tell us about missing values, skewness of data, etc.

The Reproduction tab tells us about the start and end time of the report generation, also about the duration, software version, etc. Take a look at the below image for more clearance.

Variables

This section gives us information on the variables, which tells us about the type of the variable, then distinct and missing values with the memory size that the variable is taking. Let's see the example of two variables below, id is a real number and grade is categorical.

Correlations

A statistical tool that helps in the study of the relationship between two variables is known as Correlation.

Missing Values

The profile report also gives us information on missing values in the data visually using the bar plot.

Sample

This displays the first and last 10 rows of the dataset.

Saving the Profile Report

We can save the report in either HTML or JSON format. The code for that is shown below:

# saving to html format
profile.to_file("output.html")

# saving to json format
profile.to_file("output.json")