Repository Mirror for your Cloud Server and Webhosting

Type:

Package

Title:

Randomized Feature and Bootstrap-Enhanced Gaussian Naive Bayes Classifier

Version:

0.2.4

Date:

2025-12-21

Description:

Provides an accessible and efficient implementation of a randomized feature and bootstrap-enhanced Gaussian naive Bayes classifier. The method combines stratified bootstrap resampling with random feature subsampling and aggregates predictions via posterior averaging. Support is provided for mixed-type predictors and parallel computation. Methods are described in Srisuradetchai (2025) <doi:10.3389/fdata.2025.1706417> "Posterior averaging with Gaussian naive Bayes and the R package RandomGaussianNB for big-data classification".

License:

MIT + file LICENSE

Encoding:

UTF-8

Imports:

parallel, stats

RoxygenNote:

7.3.3

Suggests:

mlbench, testthat (≥ 3.0.0)

Config/testthat/edition:

NeedsCompilation:

Packaged:

2025-12-21 16:57:28 UTC; spatc

Author:

Patchanok Srisuradetchai [aut, cre]

Maintainer:

Patchanok Srisuradetchai <patchanok@mathstat.sci.tu.ac.th>

Repository:

CRAN

Date/Publication:

2026-01-07 08:00:14 UTC

Predict from a random_gaussian_nb model

Description

Predict from a random_gaussian_nb model

Usage

## S3 method for class 'random_gaussian_nb'
predict(object, newdata = NULL, type = c("class", "prob"), ...)

Arguments

object

A fitted random_gaussian_nb object.

newdata

A data.frame of predictors. If NULL, uses training predictors.

type

"class" (default) or "prob".

...

currently unused.

Value

If type = "prob", returns a data.frame with one column per class giving posterior probabilities averaged over the bootstrap ensemble (rows correspond to observations in newdata).

If type = "class", returns a factor of predicted class labels with levels equal to the training classes.

Train a Random Naive Bayes Model via Bootstrap + Random Subspace (Mixed Types)

Description

Fits an ensemble Naive Bayes classifier by repeating (i) stratified bootstrap resampling of rows and (ii) random feature-subset selection, then aggregates predictions by posterior averaging.

Usage

## S3 method for class 'random_gaussian_nb'
print(x, ...)

## S3 method for class 'random_gaussian_nb'
summary(object, ...)

## S3 method for class 'random_gaussian_nb'
str(object, ...)

## S3 method for class 'random_gaussian_nb'
nobs(object, ...)

## S3 method for class 'random_gaussian_nb'
fitted(object, ...)

## S3 method for class 'random_gaussian_nb'
plot(
  x,
  which = c("feature_frequency", "prior_variability", "prob_entropy"),
  newdata = NULL,
  top = 20,
  ...
)

random_gaussian_nb(
  data,
  response,
  n_iter = 100,
  feature_fraction = 0.5,
  cores = 1,
  laplace = 1
)

Arguments

x

A random_gaussian_nb object.

...

Passed to the underlying plotting function (e.g., barplot(), boxplot(), hist()).

object

A random_gaussian_nb object.

which

Diagnostic to plot: "feature_frequency", "prior_variability", or "prob_entropy".

newdata

Optional new data for "prob_entropy". If NULL, uses the training data.

top

Number of top features to show for "feature_frequency".

data

A data.frame containing predictors and the response.

response

Name of the response column (string).

n_iter

Positive integer; number of bootstrap iterations.

feature_fraction

Numeric in (0,1]; fraction of features sampled each iteration.

cores

Positive integer; number of parallel workers.

laplace

Numeric >= 0; Laplace smoothing parameter for categorical features.

Details

Numeric predictors use Gaussian likelihoods; categorical predictors (factor/character/logical) use multinomial likelihoods with Laplace smoothing.

The following S3 methods are available for this class:

print(x, ...): Returns x invisibly (called for side effects).
summary(object, ...): Returns object invisibly (prints a summary).
str(object, ...): Returns object invisibly (prints a compact structure).
nobs(object, ...): Returns an integer: number of training observations.
fitted(object, ...): Returns a factor of fitted class labels for the training data.
plot(x, ...): Returns x invisibly (called for its side effects).

Value

An object of class "random_gaussian_nb" containing the fitted bootstrap ensemble and training metadata.