Feature Transformation

A mathematical transformation in which we apply a mathematical formula to a particular column(Feature) and transform it into values which are useful for further analysis.

Types of granularity in Feature Transformation:

  • Instance-level: Involves just the instance (aka one row of data).
  • Full-pass: Involves the entire dataset.

  • Feature Transformation on Categorical Variables is called a supervised feature engineering and is performed via Target Encoding task.
  • Target Encoding is an Encoding method to encode categories into numbers. It's similar to One-Hot Encoding, however it also uses the target to create the encoding.