Chapter 8: Support Vector Regression
Support vector regression (SVR) can be an excellent option when the assumptions of linear regression models do not hold, such as when the relationship between our features and our target is too complicated to be described by a linear combination of weights. Even better, SVR allows us to model that complexity without having to expand the feature space.
Support vector machines identify the hyperplane that maximizes the margin between two classes. The support vectors are the data points closest to the margin that support it, if you will. This turns out to be as useful for regression modeling as it is for classification. SVR finds the hyperplane containing the greatest number of data points. We will discuss how that works in the first section of this chapter.
Rather than minimizing the sum of the squared residuals, as ordinary least squares regression does, SVR minimizes the coefficients within an acceptable error range. Like ridge and lasso regression...
