Chapter 13: Support Vector Machine Classification

There are some similarities between support vector classification models and k-nearest neighbors models. They are both intuitive and flexible. However, support vector classification, due to the nature of the algorithm, scales better than k-nearest neighbor. Unlike logistic regression, it can handle nonlinear models rather easily. The strategies and issues with using support vector machines for classification are similar to those we discussed in Chapter 8, Support Vector Regression, when we used support vector machines for regression.

One of the key advantages of support vector classification (SVC) is the ability it gives us to reduce model complexity without increasing our feature space. But it also provides multiple levers we can adjust to limit the possibility of overfitting. We can choose a linear model or select from several nonlinear kernels. We can use a regularization parameter, much as we did for logistic regression. With...