The first time I heard of support vector machines, I have to admit that I was scratching my head, thinking that this was some form of an academic obfuscation or inside joke. However, my open-minded review of SVM has replaced this natural skepticism with a healthy respect for the technique.

SVMs have been shown to perform well in a variety of settings and are often considered one of the best "out-of-the-box" classifiers (James, G., 2013). To get a practical grasp of the subject, let's look at another simple visual example. In the following figure, you will see that the classification task is linearly separable. However, the dotted line and solid line are just two among an infinite number of possible linear solutions. You would have separating hyperplanes in a problem that has more than two dimensions:

So many solutions can be problematic for generalization because whatever solution you choose, any new observation to the right of the line will be classified as benign...