knncat (version 1.2.2)

knncat: Build a knncat classifier

Description

Build a knncat classifier, which is used for nearest-neighbor classification with categorical variables; continuous are permitted too.

Usage

knncat (train, test, k = c(1, 3, 5, 7, 9), xvals = 10, xval.ceil = -1,  knots = 10, prior.ind = 4, prior, permute = 10, permute.tail = 1,  improvement = .01, ridge = .003, once.out.always.out = FALSE,  classcol = 1, verbose = 0)

Arguments

train

data frame of training data, with the correct classification in the classcol column

test

data frame of test data (can be omitted). This should have the correct classification in the classcol column, too.

k

vector of choices for number of nn's. Default c(1, 3, 5, 7, 9).

xvals

number of cross-validations to use to find the best model size and number of nn's. Default 10.

xval.ceil

Maximum number of variables to add. -1 = Use the smallest number from any xval; 0 = use the smallest number from the first xval; >= 0, use that.

knots

vector of number of knots for numeric variables. Reused if necessary. Default: 10 for each.

prior.ind

Integer telling how to compute priors. 1 = estimated from training set; 2 = all equal; 3 = supplied in "prior"; 4 = ignored. Default: 4.

prior

Numeric vector, one entry per unique element in the training set's classcol column, giving prior probabilities. Ignored unless prior.ind = 3; then they're normalized to sum to 1 and each entry must be strictly > 0.

permute

Number of permutations for variable selection. Default: 10.

permute.tail

A variable fails the permutation test if permute.tail or more permutations do better than the original. Default: 1.

improvement

Minimum improvement for variable selection. Ignored unless present and permute missing, or permute = 0; then default = .01.

ridge

Amount by which to "ridge" the W matrix for numerical stability. Default: .003.

once.out.always.out

if TRUE, a variable that fails a permutation test or doesn't improve by enough is excluded from further consideration during that cross-validation run. Default FALSE.

classcol

Column with classification in it. Default: 1.

verbose

Controls level of diagnostic output. Higher numbers produce more output, sometimes 'way too much. 0 produces no output; 1 gives progress report for xvals. Default: 1.

Value

cdata

A vector with one entry for each of the columns of train, except the classification column, with value 1 if that column was used in the final classifier, and 0 otherwise.

phi

A list with the phi's. Each element of the list has, as its name, the name of a column of train; the values of the element are the phi's, and the names of that element are the levels of the variable. For numeric variables, these names are "knot.1", "knot.2" etc.

k

The vector of k's to be tried, as passed in.

best.k

The best k selected.

misclass.mat

A matrix, number of classes * number of classes, whose columns give the correct classifications and rows, the estimates.

prior.ind

Method used to compute the prior, as passed in.

prior

A numeric vector, one per class, giving the prior probabilties, as computed by the program according to prior.ind.

status

Return value from the program. 0 = no error.

misclass.type

Type of misclass.mat. "train" means misclass.rate came from the training set; "test," from the test set.

train

Name of training set at build time.

vars

Vector of names of columns actually used in model.

knots.vec

Vector of numbers of knots, as passed in.

build

Named vector holding five of the arguments used at build time: permute, improvement, ridge, once.out.always.out, and xvals

missing

Vector of values with which to replace missing values. These are the most common values for categorical variables, and the means for continuous ones.

knot.values

List of knot locations, one element for each continuous variable.

Details

A knncat classifier converts categorical labels into real numbers (phi) so as to produce a good k-nearest neighbor classifier. Continuous variables are handled by means of knots, in a manner similar to the linear spline representation. Variable selection is done by a permutation test, or by setting an "improvement" cutoff; error rate estimation is done by cross-validation. After the cross-validations are done, we choose the best value of k from among those proposed and the "best" number of variables, then make one more pass through all the data to estimate the phis.

References

Buttrey, S.E., Nearest-neighbor classification with categorical variables, Comp. Stat. Data Analysis 28 (1998), 157-169.

Examples

## Not run: 
# data ("synth.tr", package="MASS")
# data ("synth.te", package="MASS")
# syncat <- knncat (synth.tr, classcol=3)
# syncat
# Train set misclass rate: 12.8
# 
# synpred <- predict (syncat, synth.tr, synth.te, train.classcol=3,
#                     newdata.classcol=3)
# table (synpred, synth.te$yc)
#        
# synpred 0   1  
#       0 460  91
#       1  40 409
# #
# # Or do the whole thing in one pass:
# #
# 
# knncat (synth.tr, synth.te, classcol=3)
# Test set misclass rate: 13.1
# ## End(Not run)