Block 144: Capstone: Feature Engineering
Create meaningful features from raw data for analysis or modeling.
Concepts
- Deriving new columns from existing ones
- Encoding categorical variables
- Scaling numeric features
- Creating time-based features from datetime columns
Code Examples
See exercise below.
Exercise
Engineer at least 3 new features for your project. Justify each feature: why should it help? Run a quick correlation analysis: do your new features correlate with the target variable?
Homework
Feature engineering is often more impactful than model selection. Describe 2 domain-specific features that could improve your project significantly. Wednesday