Episode

Using Categorical Features for Better Predictions with Linear Regression (Part 13 of 17) | Machine Learning for Beginners

with Bea Stollnitz

Machine Learning for Beginners

In this video, presented by Bea Stollnitz, a Principal Cloud Advocate at Microsoft, we'll dive into categorical features, understand what they are, when to use them, and how to create them to improve the performance of our regression models.

We'll continue to work with the pumpkin dataset, which we've used in our previous videos, as we try to find the cheapest month to buy pumpkins.

In this video, you'll learn:

  • The importance of categorical features in predictions
  • How to create categorical features using pandas
  • One hot encodings
  • Improving regression results with additional features

Stay tuned for the next video in this series, where we'll look at logistic regression. See you there!

Chapters

  • 00:00 - Introduction
  • 00:33 - Recap of the code run in previous video
  • 00:50 - Can we improve our results with more features?
  • 01:16 - Pumpkin varieties as strings don't work for regression
  • 01:32 - Can we replace strings with numbers? No!
  • 01:59 - Converting strings to categorical features using one hot encoding
  • 02:40 - Improve the results using more features
  • 03:03 - Improve the results further using polynomial regression
  • This course is based on the free, open-source, 26-lesson ML For Beginners curriculum from Microsoft.
  • The Jupyter Notebook to follow along with this lesson is available!

Connect

Azure Machine Learning
Python