predict.ERWNN: Predicting targets of an ERWNN-object

In svilsen/BRVFL: Random Weight Neural Networks

Predicting targets of an ERWNN-object

Description

Predicting targets of an ERWNN-object

Usage

## S3 method for class 'ERWNN'
predict(object, ...)

Arguments

object	An ERWNN-object.
...	Additional arguments.

Details

The additional arguments 'newdata', 'type', and 'class' can be specified as follows:

newdata

Expects a matrix or data.frame with the same features (columns) as in the original data.

type

A string taking the following values:

"mean" (default)

Returns the average prediction across all ensemble models.

"std"

Returns the standard deviation of the predictions across all ensemble models.

"all"

Returns all predictions for each ensemble models.

class

A string taking the following values:

"classify"

Returns the predicted class of the ensemble. If used together with type = "mean", the average prediction across the ensemble models are used to create the classification. However, if used with type = "all", every ensemble is classified and returned.

"voting"

Returns the predicted class of the ensemble by classifying each ensemble and using majority voting to make the final prediction. NB: the type argument is overruled.

Furthermore, if 'class' is set to either "classify" or "voting", additional arguments 't' and 'b' can be passed to the classify-function.

NB: if the ensemble is created using the boost_rwnn-function, then type should always be set to "mean".

Value

An list, matrix, or vector of predicted values depended on the arguments 'method', 'type', and 'class'.

svilsen/BRVFL documentation built on Feb. 26, 2025, 7:43 a.m.