Support Vector Regression

Support Vector Regression (SVR) finds a function that stays within an epsilon-margin of the true values for most training points. It is effective for nonlinear regression with kernel transformations.

When to use:

• Small-to-medium datasets with nonlinear feature-target relationships
• When a robust fit within a tolerance band is desired
• High-dimensional feature spaces

Input: Tabular data with the feature columns defined during training Output: Continuous predicted value

Model Settings (set during training, used at inference)

Kernel (default: rbf) Kernel function. rbf is the standard nonlinear choice; linear for large sparse datasets.

C (default: 1.0) Penalty for points outside the epsilon tube. Higher C fits more closely.

Epsilon (default: 0.1) Width of the no-penalty tube around the prediction. Larger values create smoother models.

Gamma (default: scale) Kernel coefficient for rbf and poly. Lower values give broader influence.

Degree (default: 3) Degree for the poly kernel.

Inference Settings

No dedicated inference-time settings. The trained support vectors define the regression function.