Skip to main content
SVR finds a function that fits as many data points as possible within an epsilon-wide tube around the prediction. Points outside the tube are penalized.

​
Configuration

ParameterDescriptionDefault
CRegularization parameter. Higher = less regularization, tighter fit.1.0
KernelKernel function: rbf, linear, poly, sigmoidrbf
GammaKernel coefficient: scale, auto, or a float.scale
EpsilonWidth of the insensitive tube. Predictions within ε of the true value incur no penalty.0.1

​
Input / Output

Type
InputX Train + Y Train
OutputTrained Model

​
When to use

Effective on small-to-medium datasets with non-linear relationships. Always scale features before SVR.