Package 'ggsom'

Visualization in parallelels coordinates in matrix of each attribute

Description

Visualization of the classes corresponding to each neuron of the SOM

Usage

geom_class(object_som, class = NULL, x_o = 3, y_o = 5.5, x_e = 3, y_e = 6.3)

Arguments

object_som	object of Kohonen package
class	categorical vector corresponding to the class of the dataset
x_o	x-axis to map the number of observations of each neuron
y_o	y-axis to map the number of observations of each neuron
x_e	x-axis to map the entropy of each neuron
y_e	y-axis to map the entropy of each neuron

Value

ggplot2 object

Author(s)

Felipe Carvalho, [email protected]

References

'ggplot2' package (https://CRAN.R-project.org/package=ggplot2)

Examples

# Creating SOM object
iris_som <- kohonen::som(X = as.matrix(iris[1:4]),
                         grid =  kohonen::somgrid(xdim = 5,
                                                   ydim = 5,
                                                   neighbourhood.fct = "gaussian",
                                                   topo = "rectangular"),
                         rlen = 100)

# Creating ggsom class plot
geom_class(iris_som, class = iris$Species,
           x_o = 1, y_o = 6,
           x_e = 1.1, y_e = 7.4)

---

ggsom

Description

The aim of this package is to offer more variability of graphics based on the self-organizing maps

---

kohonen package object modeling

Description

Function to map each SOM neuron with its corresponding class

Usage

ggsom_aes(object_som, class)

Arguments

object_som	object of kohonen package
class	categorical vector corresponding to the class of the dataset

Value

data.table model used in visualizations

Author(s)

Felipe Carvalho, [email protected]

References

'Kohonen'package (https://CRAN.R-project.org/package=kohonen)

---

Function to obtain the purity of each neuron in the SOM network

Description

Entropy calculation using the maximum likelihood method

Usage

ggsom_entropy(ggsom_aes)

Arguments

ggsom_aes

kohonen package object modeling

Value

Data set with the purity attribute added in Tibble

Author(s)

Felipe Carvalho, [email protected]

---

verifies that the object inherits kohonen object

Description

if object inherits kohonen class return TRUE otherwise FALSE

Usage

is.kohonen(object_som)

Arguments

object_som

object of Kohonen package

Value

Boolean value

Author(s)

Felipe Carvalho, [email protected]

References

'Kohonen'package (https://CRAN.R-project.org/package=kohonen)

---