juggling {cccd}

Juggling

Description

a resampled version of the CCCD classifier.

Usage

juggle(data, classes, sampled = TRUE, sample.dim = FALSE, 
       num = 100, sample.proportion = 0.1, k = 2, method = NULL)
juggle.classify(data,J,tdata,indices)

Arguments

data, tdata

training data from which to build the classifier. In the case of juggle.classify, tdata is the training data and data is the test data.

classes

class labels.

sampled

whether the data are subsampled.

sample.dim

if TRUE, the dimensions (variates) are also sampled.

num

number of juggles (resamples).

sample.proportion

proportion of the data to sample. If 1 or greater, the data are sampled with replacement.

k

number of variates to sample when sample.dim is TRUE.

J

the juggled classifier.

indices

the indices of the juggles to use.

method

the method used for the distance. See dist

Details

The idea of juggling is to sample the data, compute a CCCD classifier, then repeat. The resampling is controled by the two sampling variables, which basically determine whether the data are sampled with replacement, or whether a subsample is used. If sample.dim is TRUE, the variates are also sampled, with k indicating how many are sampled.

Value

juggle.classify returns a matrix holding the classification probabilities for each observation in data. a list consisting of:

S

the dominating sets.

R

the radii.

dimension

the dimension of the data.

vars

in the case of sample.dim=TRUE, the variables sampled each time.

Only the indicies into the training data are stored in J, which is why the classifier requires the original training data in tdata.

Author(s)

David J. Marchette, [e-mail]

See Also

cccd, dist