Introduction to the noisemodel package

The hardware and bandwidth for this mirror is donated by METANET, the Webhosting and Full Service-Cloud Provider.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]metanet.ch.

The noisemodel package contains the first extensive implementation of noise models for classification datasets. It provides 72 noise models found in the specialized literature that allow errors to be introduced in different ways in class labels, attributes or both in combination. Each of them is properly documented and referenced, unifying their results through a specific S3 class, which benefits from customized print, summary and plot methods.

Installation

The noisemodel package can be installed in R from CRAN servers using the command:

# install.packages("noisemodel")

This command installs all the dependencies of the package that are necessary for the operation of the noise models. In order to access all the functions of the package, it is necessary to use the R command:

library(noisemodel)

Documentation

All the information corresponding to each noise model can be consulted from the CRAN website. Additionally, the help() command can be used. For example, in order to check the documentation of the Symmetric uniform label noise model, we can use the command:

# help(sym_uni_ln)

Usage of noise models

For introducing noise in a dataset, each noise model in the noisemodel package provides two standard ways of use:

Default method. It receives a data frame with the input attributes in the x argument, whereas the output variable is received through the y argument (a factor vector).
Class formula method. This method allows passing the whole data frame (attributes and response variable) in the data argument. In addition, the attributes along with the output label must be indicated in the formula argument.

An example on how to use these two methods for introducing noise in the iris2D dataset with the sym_uni_ln model is shown below:

# load the dataset
data(iris2D)

# usage of the default method
set.seed(9)
outdef <- sym_uni_ln(x = iris2D[,-ncol(iris2D)], y = iris2D[,ncol(iris2D)], level = 0.1)

# show results
summary(outdef, showid = TRUE)
#> 
#> ########################################################
#>  Noise introduction process: Summary
#> ########################################################
#> 
#> ## Original call:
#> sym_uni_ln(x = iris2D[, -ncol(iris2D)], y = iris2D[, ncol(iris2D)], level = 0.1)
#> 
#> ## Noise model: 
#> Symmetric uniform label noise
#> 
#> ## Parameters:
#> - level = 0.1
#> - sortid = TRUE
#> 
#> ## Number of noisy and clean samples:
#> - Noisy samples: 10/101 (9.90%)
#> - Clean samples: 91/101 (90.10%)
#> 
#> ## Number of noisy samples per class label:
#> - Class setosa: 3/22 (13.64%)
#> - Class versicolor: 4/35 (11.43%)
#> - Class virginica: 3/44 (6.82%)
#> 
#> ## Number of clean samples per class label:
#> - Class setosa: 19/22 (86.36%)
#> - Class versicolor: 31/35 (88.57%)
#> - Class virginica: 41/44 (93.18%)
#> 
#> ## Indices of noisy samples:
#> - Output class: 3, 6, 12, 24, 30, 48, 53, 59, 83, 101
plot(outdef)


# usage of the method for class formula
set.seed(9)
outfrm <- sym_uni_ln(formula = Species ~ ., data = iris2D, level = 0.1)

# check the match of noisy indices
identical(outdef$idnoise, outfrm$idnoise)
#> [1] TRUE

Note that, the $ operator is used to access the elements returned by the noise model in the objects outdef and outfrm.

Output values

All noise models return an object of class ndmodel. It is designed to unify the output value of the methods included in the noisemodel package. The class ndmodel is a list of elements with the most relevant information of the noise introduction process:

xnoise a data frame with the noisy input attributes.
ynoise a factor vector with the noisy output class.
numnoise an integer vector with the amount of noisy samples per class.
idnoise an integer vector list with the indices of noisy samples.
numclean an integer vector with the amount of clean samples per class.
idclean an integer vector list with the indices of clean samples.
distr an integer vector with the samples per class in the original data.
model the full name of the noise introduction model used.
param a list of the argument values.
call the function call.

As an example, the structure of the ndmodel object returned using the sym_uni_ln model is shown below:

str(outdef)
#> List of 10
#>  $ xnoise  :'data.frame':    101 obs. of  2 variables:
#>   ..$ Petal.Length: num [1:101] 1.4 1.3 1.5 1.7 1.4 1.5 1.6 1.4 1.1 1.2 ...
#>   ..$ Petal.Width : num [1:101] 0.2 0.2 0.2 0.4 0.3 0.1 0.2 0.1 0.1 0.2 ...
#>  $ ynoise  : Factor w/ 3 levels "setosa","versicolor",..: 1 1 2 1 1 2 1 1 1 1 ...
#>  $ numnoise: Named int [1:3] 3 4 3
#>   ..- attr(*, "names")= chr [1:3] "setosa" "versicolor" "virginica"
#>  $ idnoise :List of 1
#>   ..$ : int [1:10] 3 6 12 24 30 48 53 59 83 101
#>  $ numclean: Named int [1:3] 19 31 41
#>   ..- attr(*, "names")= chr [1:3] "setosa" "versicolor" "virginica"
#>  $ idclean :List of 1
#>   ..$ : int [1:91] 1 2 4 5 7 8 9 10 11 13 ...
#>  $ distr   : Named int [1:3] 22 35 44
#>   ..- attr(*, "names")= chr [1:3] "setosa" "versicolor" "virginica"
#>  $ model   : chr "Symmetric uniform label noise"
#>  $ param   :List of 2
#>   ..$ level : num 0.1
#>   ..$ sortid: logi TRUE
#>  $ call    : language sym_uni_ln(x = iris2D[, -ncol(iris2D)], y = iris2D[, ncol(iris2D)], level = 0.1)
#>  - attr(*, "class")= chr "ndmodel"

In order to display the results of the class ndmodel in a friendly way in the R console, specific print, summary and plot functions are implemented. The print function presents the basic information about the noise introduction process contained in an object of class ndmodel:

print(outdef)
#> 
#> ## Noise model: 
#> Symmetric uniform label noise
#> 
#> ## Parameters:
#> - level = 0.1
#> - sortid = TRUE
#> 
#> ## Number of noisy and clean samples:
#> - Noisy samples: 10/101 (9.90%)
#> - Clean samples: 91/101 (90.10%)

The information offered by print is as follows:

the name of the noise introduction model.
the parameters associated with the noise model.
the number of noisy and clean samples in the dataset.

On the other hand, the summary function displays a summary containing information about the noise introduction process contained in an object of class ndmodel, with other additional details. This function can be called by typing the following R command:

summary(outdef, showid = TRUE)
#> 
#> ########################################################
#>  Noise introduction process: Summary
#> ########################################################
#> 
#> ## Original call:
#> sym_uni_ln(x = iris2D[, -ncol(iris2D)], y = iris2D[, ncol(iris2D)], level = 0.1)
#> 
#> ## Noise model: 
#> Symmetric uniform label noise
#> 
#> ## Parameters:
#> - level = 0.1
#> - sortid = TRUE
#> 
#> ## Number of noisy and clean samples:
#> - Noisy samples: 10/101 (9.90%)
#> - Clean samples: 91/101 (90.10%)
#> 
#> ## Number of noisy samples per class label:
#> - Class setosa: 3/22 (13.64%)
#> - Class versicolor: 4/35 (11.43%)
#> - Class virginica: 3/44 (6.82%)
#> 
#> ## Number of clean samples per class label:
#> - Class setosa: 19/22 (86.36%)
#> - Class versicolor: 31/35 (88.57%)
#> - Class virginica: 41/44 (93.18%)
#> 
#> ## Indices of noisy samples:
#> - Output class: 3, 6, 12, 24, 30, 48, 53, 59, 83, 101

The information offered by this function is as follows:

the function call.
the name of the noise introduction model.
the parameters associated with the noise model.
the number of noisy and clean samples in the dataset.
the number of noisy samples per class/attribute.
the number of clean samples per class/attribute.
the indices of the noisy samples (if showid = TRUE).

Finally, the plot function displays a representation of the dataset contained in an object of class ndmodel after the application of a noise introduction model.

plot(outdef)

This function performs a two-dimensional representation using the ggplot2 package of the dataset contained in the object x of class ndmodel. Each of the classes in the dataset (available in x$ynoise) is represented by a different color.

These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.