predict.LogitBoost: Prediction Based on LogitBoost Classification Algorithm
Description
Prediction or Testing using logitboost classification algorithmUsage
"predict"(object, xtest, type = c("class", "raw"), nIter=NA, ...)
Arguments
object
An object of class "LogitBoost" see "Value" section of
LogitBoost for details xtest
A matrix or data frame with test data. Rows contain samples
and columns contain features
nIter
An optional integer, used to lower number of iterations
(decision stumps) used in the decision making. If not provided than the
number will be the same as the one provided in LogitBoost.
If provided than the results will be the same as running
LogitBoost with fewer iterations. ...
not used but needed for compatibility with generic predict
method
Value
If type = "class" (default) label of the class with maximal probability is
returned for each sample. If type = "raw", the a-posterior probabilities for
each class are returned.
Details
Logitboost algorithm relies on a voting scheme to make classifications. Many
(nIter of them) week classifiers are applied to each sample and their
findings are used as votes to make the final classification. The class with
the most votes "wins". However, with this scheme it is common for two cases
have a tie (the same number of votes), especially if number of iterations is
even. In that case NA is returned, instead of a label.
See Also
LogitBoost has training half of LogitBoost code