Data Science Desktop Survival Guide
by Graham Williams
One-Class Support Vector Machine
|Representation:||Bounding Support Vectors|
A support vector machine (Section 16.18) identifies a linear separator between observations that belong to different classes. If the observations are not tagged with a class then the algorithm can be utilised to identify a bounding region for the training dataset. The model is then represented as the boundary support vectors.
No all observations from the training dataset will sit within the boundary. That is, not all observations will fit the model. Such observations we might consider as outliers.
Explore one-class support vector machines through the ocsvm package from MLHub.