Feature Engineering in Machine Learning Pipelines

Feature engineering plays a significant role in building high-performing machine learning models. It’s about creating new signals that better represent the underlying patterns in the domain you’re analyzing. In this lesson, we’ll explore how to thoughtfully design features using domain expertise and apply common transformation techniques like encoding, interaction terms, and discretization to prepare customer data for modeling. Let’s get started.

Creating a feature effectively

You’re interviewing with a company that partners with retailers to optimize their operations. How would you use feature engineering to create a new feature, leveraging their retail partners’ domain expertise? You may also receive a follow-up question along the lines of: What kind of data would you need to create this feature?

Sample answer

Here’s a viable answer to this question, focusing on the company’s domain:

I would use feature engineering to create a new feature called Seasonal Demand Index. Using domain expertise, I’d analyze historical sales data to identify patterns in customer purchases throughout different seasons. For example, winter clothing may sell more in ...