Jupyter EDA is a small set of tools designed to assist with exploratory data analysis (EDA) in Jupyter Notebook by wrapping around and providing convenient access to tools for data munging, analysis, and visualization from modules such as:

• Matplotlib (for data visualization)
• NumPy (for data handling and manipulation)
• Pandas (for data handling and manipulation)
• SciKit-learn (for data handling and manipulation)
• Seaborn (for data visualization)

It also relies on a few other modules for reporting and functionality:

• IPython (for output formatting in Jupyter Notebook)
• Jinja2 (for HTML and CSS templating)
• TQDM (for graphical progress bars)
• Weasyprint (for PDF output)

We all know that before you can start loading your data into statistical models for machine learning, there is a lot of cleaning and general data "munging" or "wrangling" that has to be done so that your dataset is ready to start throwing linear algebra at it. Unfortunately, despite the power of tools like NumPy, Pandas, and SciKit-learn, it can be difficult to really get down and dirty with the data while also ensuring process reproducibility and repeatability. With that in mind, Jupyter EDA focuses on providing four kinds of functionality to speed up and improve your EDA and data preprocessing workflows:

1. Tracking
   • Organize and track your data to ensure you know what you're working with, when it was edited, and how
   • Log your cleaning and feature engineering process to ensure reproducibility
   • Segregate training/testing data and different variable types to ensure proper data handling
2. Analyzing
   • Perform a variety of automated initial data analysis (IDA) and exploratory data analysis (EDA) functions
   • Use special analysis tools for data such as text and datetime objects to examine useful patterns prior to feature engineering
3. Producing
   • Use automated functions to clean your data and impute missing values based on various best practices
   • Generate pre-designed reports and plots to get a snapshot of key aspects of your dataset
   • Engineer new features using built-in or ad hoc "recipes"
4. Exporting
   • Export your data, logs, or even a whole pickled Dataset object in case you need to move your work somewhere else
   • Save your reports and visualizations in a variety of useful formats, including PDF and embed-ready HTML
   • Repackage your data preprocessing steps as pipeline-ready objects with fit() and transform() functions
   • Create custom data objects that are ready for monitoring