Feature engineering

MLOps Concepts

Folkert Stijnman

ML Engineer

Feature engineering

MLOps phases feature engineering

MLOps Concepts

Feature engineering

... is the process of selecting, manipulating, and transforming raw data into features.

  • A feature is a variable, such as the column in a table
  • We can use raw data, but also create our own
MLOps Concepts

Customer data

Example customer data

MLOps Concepts

Customer data

Example customer data with new feature

MLOps Concepts

Feature engineering

  • Goal is to enhance model performance
  • Tools and techniques help to process, select, and maintain features:
    • Feature selection
    • Feature store
    • Data version control
MLOps Concepts

Feature selection

  • Domain-specific knowledge
  • Correlation
  • Feature importances
  • Other methods: univariate selection, Principal Component Analysis (PCA), Recursive Feature Elimination (RFE)

correlation plot

1 https://www.datacamp.com/tutorial/tutorial-datails-on-correlation
MLOps Concepts

The feature store

Feature store

Only relevant for large teams working on multiple projects that use the same features

MLOps Concepts

Data version control

  • Tracking dataset changes
  • Maintaining consistency throughout the development lifecycle

graphic depicting data version control and git

1 https://www.datacamp.com/courses/cicd-for-machine-learning
MLOps Concepts

Let's practice!

MLOps Concepts

Preparing Video For Download...