ada (version 2.0-5)

predict.ada: Predict a data set using Ada

Description

predict classifies a new set of observations from a previously built classifier. This function will provide either a vector of new classes, class probability estimates, or both.

Usage

"predict"(object, newdata, type = c("vector", "probs", "both","F"),n.iter=NULL,...)

Arguments

object

object generated by ada.

newdata

new data set to predict. This data set must be of type ‘data.frame’ and prediction data set is required for this approach.

type

choice for preditions. type=“vector” returns the default class labels. type=“prob” returns the probability class estimates. type=“both” returns both the default class labels and probability class estimates. type=“F” returns the ensamble average, where the class label is sign(F). This is mainly usefull for the multiclass case.

n.iter

number of iterations to consider for the prediction. By default this is iter from the ada call (n.iter< iter)

...

other arguments not used by this function.

Value

fit

a vector of fitted responses. Fit will be returned if type=“vector”.

probs

a matrix of class probability estimates. The first column corresponds to the first label in the ‘levels’ of the response. The second column corresponds to the second label in the ‘levels’ of the response. Probs are returned whenever type=“probs”.

both

returns both the vector of fitted responses and class probability estimates. The first element returns the fitted responses and will be labeled as ‘class’. The second element returns the class probability estimates and will be labeled as ‘probs’.

F

this is used in the multiclass case when one uses the package to perform 1 v.s. all.

Details

This function was modeled after predict.rpart. Furthermore, predict.rpart will be invoked to handle predictions by each tree in the ensamble.

See Also

ada.default,summary.ada,print.ada, plot.ada,pairs.ada,update.ada,addtest