Huber Regressor

Linear regression that's robust to outliers by using a combination of squared and absolute loss.

When to use:

• Have outliers in target variable
• Want linear model but data has outliers
• Need robust predictions
• Alternative to removing outliers

Strengths: Robust to outliers, maintains linear interpretability, handles noisy data Weaknesses: Slower than ordinary linear regression, one extra hyperparameter

Model Parameters

Epsilon (default: 1.35) Threshold where loss changes from squared to linear. Smaller = more outliers treated as outliers.

Alpha (default: 0.0001) Regularization parameter.

Max Iterations (default: 100) Maximum optimization iterations.

Fit Intercept (default: true) Whether to calculate intercept term.