R Statistics Cookbook

How it works...

SVMs can be used for both regression and classification. The underlying idea is to generate a sequence of separating hyperplanes that separate the data as efficiently as possible. Conceptually, a separating hyperplane can be thought of as drawing a line on the floor that separates what is covered by a carpet, and where the furniture is. Of course, there will be overlaps (furniture sitting on top of the carpet), and the furniture might also be very close to the carpet. This defines a triple problem:

• The margin separating furniture from the carpet can be thin.
• We will usually have an overlap.
• The hyperplane separating the furniture and the carpet might not necessarily be linear; for example, a curve might separate them optimally. ...