Forecaster for SVM Model Fits

This function predicts values based on Support Vector Machines model objects.

Usage

forecast_by_svm(
  object,
  newx,
  type = c("response", "class"),
  newoffset,
  pthr = 0.5,
  ...
)

object

an object of class Trained

newx

the new values at which prediction are to be made

type

character specifying the type of prediction. Three options are available

"response": predicted values are on the same scale as the outcome, e.g. for binomial the returned values are classes, for gaussian the expected values
"class": the class corresponding to the max probability

newoffset

currently not used

pthr

probability threshold, used when type = "class" and response is binomial to assign label to observation

...

further arguments to predict.svm

A vector containing the predictions.

It seems there is an issue with the underlying function when asking for probabilities

Alessandro Barberis