FeatureHub

FeatureHub is a web application built on top of JupyterHub and an accompanying Python package that together facilitate collaborative data science efforts as well as data collection experiments on those efforts.

The FeatureHub Python package is used to administer a data science collaboration. It allows users to “register” features in a database backend and discover features written by other users. These features are automatically scored by the system so that users can see realtime feedback on the quality of their features.

The FeatureHub app includes a JupyterHub server, containerized Jupyter notebooks for each user with the FeatureHub package installed, a database backend to store the features, and data- and user-management tools.

Table of Contents

• Intro
• User Guide
  • Accessing FeatureHub
  • Accessing FeatureHub Discourse forum
  • Using Jupyter Notebook
• FeatureHub Tutorial
  • Outline
  • Prepare your session
  • Acquire dataset
  • Explore existing features
  • Write a new feature
  • Evaluate a feature on training data
  • Submit a feature to Feature database
  • Final notes
  • More feature scaffolding
  • Getting help
• FAQ
  • How do I restart my FeatureHub server?
  • Why is my feature rejected by the submit function?
  • How can I see every column of a wide DataFrame?
• Deployment
  • Installation
  • Configuration
  • Data management
  • User management
  • App administration
  • Manage app lifecycle
  • Other topics

Indices and tables

• Index
• Module Index
• Search Page