Support Vector Machine (SVM) is a powerful and advanced supervised learning technique for classification and regression that can automatically fit linear and nonlinear models.
SVM algorithms have quite a few advantages against other machine learning algorithms:
- They can handle the majority of supervised problems such as regression, classification, and anomaly detection (anyway, they are actually best at binary classification).
- They provide a good handling of noisy data and outliers. They tend to overfit less, since they only work with some particular examples, the support vectors.
- They work fine with datasets presenting more features than examples, though, as with other machine learning algorithms, SVM would gain both from dimensionality reduction and feature selection.
- As for drawbacks, we have to mention these:
- They provide only estimates, but no probabilities...