Introduction
SVMs can also predict numbers. The trick is to reverse the classification goal:
- SVM classification: fit the widest street that keeps the classes outside it.
- SVM regression (SVR): fit a street (a tube) that contains as many points as possible inside it.
Think of laying a pipe of fixed width along a scattered row of trees: you want as many trees as possible to fit inside the pipe, and you only worry about the trees sticking out.