Polynomial Regression

Extends linear regression by creating polynomial feature combinations to capture non-linear relationships.

When to use:

• Relationships are curved, not straight lines
• Have few features (exponential feature growth)
• Need interpretable non-linear model
• Alternative to tree-based for small data

Strengths: Captures non-linearity while staying interpretable, flexible curve fitting Weaknesses: Exponential feature growth, prone to overfitting at high degrees, extrapolates poorly

Model Parameters

Degree (default: 2) Polynomial degree:

• 2: Quadratic (curves)
• 3: Cubic (more complex curves)
• 4+: Very complex (high risk of overfitting)

Interaction Only (default: false) If true, only feature interactions (no powers). Reduces features.

Include Bias (default: true) Whether to include constant term.