SVMACHINES: Stata module providing Support Vector Machines for both Classification and Regression

svmachines fits a support vector machine (SVM) model. SVM is not one, but several, variant models each based upon the principles of splitting hyperplanes and the culling of unimportant observations. The basic SVM idea is to find a linear boundary--a hyperplane--in high-dimensional space: for classification, this is a boundary between two classes; for regression it is a line near which points should be--much like in OLS, while simultaneously minimizing the number of observations required to distinguish this hyperplane. The unimportant observations are ignored after fitting is done, which makes SVM very memory efficient. Each observation can be thought of as a vector, so the support vectors are those observations which the algorithm deems critical to the fit.