YData Profiling

YData Profiling (formerly Pandas Profiling) is a tool that automatically generates detailed exploratory data analysis (EDA) reports from a pandas DataFrame.

It provides summaries such as:

• Data types and missing values
• Descriptive statistics
• Correlations between features
• Duplicates, distributions, and warnings
• Interactive visualisations

It’s a great first step after loading a dataset into pandas — saving you hours of manual exploration.

Installation

You can install it via pip:

pip install ydata-profiling

If you’re using Jupyter notebooks, restart the kernel after installing.

Basic Usage with Pandas

import pandas as pdfrom ydata_profiling import ProfileReport​# Example datasetdf = pd.read_csv("data.csv")​# Create a basic profile reportprofile = ProfileReport(df)​# for Jupyter displayprofile.to_notebook_iframe()   

Alternatively, export to an HTML file:

profile.to_file("data_profile.html")

Key Options in ProfileReport

• minimal - Creates a lighter, faster report (no correlations, fewer visuals) eg. ProfileReport(df, minimal=True)
• explorative - Adds interactive, advanced visuals eg. ProfileReport(df, explorative=True)
• title - Sets a custom report title eg. ProfileReport(df, title="Customer Data Profiling")
• samples - Number of samples shown per column eg. ProfileReport(df, samples={"head": 5, "tail": 5})
• correlations - Controls correlation methods eg. ProfileReport(df, correlations={"pearson": {"calculate": True}})
• config_file - Load a YAML config for advanced setups eg. ProfileReport(df, config_file="profile_config.yaml")

Typical Workflow

import pandas as pdfrom ydata_profiling import ProfileReport​# 1. Load your datadf = pd.read_csv("dataset.csv")​# 2. Inspect data quicklyprint(df.info())print(df.describe())​# 3. Generate the profiling reportprofile = ProfileReport(df, title="EDA Report", explorative=True)​# 4. Display in notebook or export Inline in Jupyterprofile.to_notebook_iframe()  ​# or Export to file# profile.to_file("report.html")  

When to Use

Use YData Profiling right after loading — it’s ideal for:

• Understanding dataset structure
• Detecting missing or inconsistent data
• Spotting potential outliers
• Getting a quick sense of correlations or redundancy