Type: Package
Title: An Implementation of Interaction Graphs of Aleks Jakulin
Version: 1.0.0
Description: Generates a 'Graphviz' graph of the most significant 3-way interaction gains (i.e. conditional information gains) based on a provided discrete data frame. Various output formats are supported ('Graphviz', SVG, PNG, PDF, PS). For references, see the webpage of Aleks Jakulin http://stat.columbia.edu/~jakulin/Int/.
Depends: R (≥ 3.5.0), dplyr (≥ 0.7.6), DiagrammeR (≥ 1.0.0), DiagrammeRsvg (≥ 0.1), rsvg (≥ 1.3), gtools (≥ 3.5.0), utils (≥ 3.5.0)
License: GPL-3
Encoding: UTF-8
LazyData: true
URL: https://github.com/peleplay/integr
BugReports: https://github.com/peleplay/integr/issues
RoxygenNote: 6.1.1
Suggests: knitr, rmarkdown, testthat
VignetteBuilder: knitr
NeedsCompilation: no
Packaged: 2019-05-24 11:14:53 UTC; peleplay
Author: Petar Markovic [aut, cre]
Maintainer: Petar Markovic <peleplay@gmail.com>
Repository: CRAN
Date/Publication: 2019-05-24 11:50:03 UTC

Calculates Shannon's entropy

Description

Formula: H(S) = -P_i * \sum log_2 * P_i, where P_i is the probability of the corresponding i-th class

Usage

entropy(df, classAtt)

Arguments

df

A discrete data.frame

classAtt

A class column of the df (string)

Value

The Shannon's entropy of the df, based on the classAtt attribute

Examples

entropy(golf, "Play")

Golf example dataset for Interaction graphs

Description

An example dataset containing the discrete data.frame (i.e. all columns are factors) with variables used as an input for making a decision whether a party of golf would be played, or not.

Usage

golf

Format

A data.frame with 6 discrete variables (i.e. factors) and 14 rows (i.e. observations). 5 input variables and 1 class (i.e. context) variable:

Outlook

Input attribute, values: Overcast, Rainy, Sunny

Temperature

Input attribute, values: Cool, Hot, Mild

Humidity

Input attribute, values: High, Normal

Windy

Input attribute, values: True, False

Others

Artifically added input attribute indicating whether the players on the other courts were playing the golf at the given time, values: Yes, No

Play

Class attribute, indicating whether the decision was to play or not to play a party of golf, values: Yes, No

@source https://gerardnico.com/data_mining/weather

Constructs Interaction Graph (S3 class)

Description

Constructs Interaction Graph (S3 class)

Usage

ig(n, e)

Arguments

n

ig.nodes (a list of igNode objects)

e

ig.edges (a list of igEdge objects)

Value

An instance of the ig class

Constructs Interaction Graph Edges (S3 class)

Description

Constructs Interaction Graph Edges (S3 class)

Usage

igEdge(n1, n2, w)

Arguments

n1

igEdge.node1 (character)

n2

igEdge.node2 (character)

w

igEdge.weight (i.e. 3-way Interaction Gain) (double)

Value

An instance of the igEdge class

Constructs Interaction Graph Nodes (S3 class)

Description

Constructs Interaction Graph Nodes (S3 class)

Usage

igNode(n, v)

Arguments

n

igNode.name (character)

v

igNode.value (double) (i.e. 2-way Interaction Gain)

Value

An instance of the igNode class

Exports Interaction graph to a GraphViz file

Description

Exports Interaction graph to a GraphViz file

Usage

igToGrViz(ig, path = "", fName = "InteractionGraph")

Arguments

ig

Interaction graph

path

The folder in which to write the GraphViz file;

fName

The name of the file to be created; "InteractionGraph" by default

Value

Writes the ig interaction graph to a GraphViz .gv file to the folder specified in the path

Examples

#create temp dir path with slashes
myDir <- gsub("\\\\", "/", tempdir())

#create interaction graph
g <- interactionGraph(golf, "Play", intNo = 10)

#write to 'graphviz' file
igToGrViz(g, path = myDir, fName = "MyGraph")

Exports Interaction graph to a PDF file

Description

Exports Interaction graph to a PDF file

Usage

igToPDF(ig, path = "", fName = "InteractionGraph", h = 2000)

Arguments

ig

Interaction graph

path

The folder in which to write the PDF file;

fName

The name of the file to be created; "InteractionGraph" by default

h

Desired height of the image in pixels; 2000px by default

Value

Writes the ig interaction graph to a PDF (.pdf) file to the folder specified in the path

Examples

#create temp dir path with slashes
myDir <- gsub("\\\\", "/", tempdir())

#create interaction graph
g <- interactionGraph(golf, "Play", intNo = 10)

#write to PDF
igToPDF(g, path = myDir, fName = "MyGraph", h = 2000)

Exports Interaction graph to a PNG file

Description

Exports Interaction graph to a PNG file

Usage

igToPNG(ig, path = "", fName = "InteractionGraph", h = 2000)

Arguments

ig

Interaction graph

path

The folder in which to write the PNG file;

fName

The name of the file to be created; "InteractionGraph" by default

h

Desired height of the image in pixels; 2000px by default

Value

Writes the ig interaction graph to a PNG (.png) file to the folder specified in the path

Examples

#create temp dir path with slashes
myDir <- gsub("\\\\", "/", tempdir())

#create interaction graph
g <- interactionGraph(golf, "Play", intNo = 10)

#write to PNG
igToPNG(g, path = myDir, fName = "MyGraph", h = 2000)

Exports Interaction graph to a PS (PostScript) file

Description

Exports Interaction graph to a PS (PostScript) file

Usage

igToPS(ig, path = "", fName = "InteractionGraph", h = 2000)

Arguments

ig

Interaction graph

path

The folder in which to write the PS file;

fName

The name of the file to be created; "InteractionGraph" by default

h

Desired height of the image in pixels; 2000px by default

Value

Writes the ig interaction graph to a PostScript (.ps) file to the folder specified in the path

Examples


#create temp dir path with slashes
myDir <- gsub("\\\\", "/", tempdir())

#create interaction graph
g <- interactionGraph(golf, "Play", intNo = 10)

#write to PS
igToPS(g, path = myDir, fName = "MyGraph", h = 2000)

Exports Interaction graph to a SVG file

Description

Exports Interaction graph to a SVG file

Usage

igToSVG(ig, path = "", fName = "InteractionGraph", h = 2000)

Arguments

ig

Interaction graph

path

The folder in which to write the SVG file;

fName

The name of the file to be created; "InteractionGraph" by default

h

Desired height of the image in pixels; 2000px by default

Value

Writes the ig interaction graph to a SVG (.svg) file to the folder specified in the path

Examples

#create temp dir path with slashes
myDir <- gsub("\\\\", "/", tempdir())

#create interaction graph
g <- interactionGraph(golf, "Play", intNo = 10)

#write to SVG
igToSVG(g, path = myDir, fName = "MyGraph", h = 2000)

Calculates Information Gain (2-way Interaction Gain) of a discrete data.frame

Description

InfoGAIN = H(S) - H(S|X), where H(S) is the difference in the Shannon's entropy of the system S before a new attribute X is introduced, and H(S|X) is the entropy of the system after the attribute X has been introduced.

Usage

infoGain(df, inAtt, classAtt)

Arguments

df

A discrete data.frame

inAtt

An input column of the data.frame df (string)

classAtt

A class column of the data.frame df (string)

Value

The Information Gain of df on the class attribute classAtt

Examples

infoGain(golf, "Windy", "Play")

infoGain(golf, "Outlook", "Play")

Creates Interaction graph

Description

Creates Interaction graph

Usage

interactionGraph(df, classAtt, intNo = 16, speedUp = FALSE)

Arguments

df

A discrete data.frame

classAtt

A class column of the df (string)

intNo

A desired number of interactions to show, i.e. an (integer) in range: [2,20]; Default value is 16.

speedUp

A (boolean) parameter. If TRUE, indicates whether the pairs of attributes with Information Gain equal to zero (on the 4th decimal) should be pruned. This speeds up calculations for larger datasets. By default it is turned off (i.e. set to FALSE).

Value

An interaction graph object (string)

Examples

interactionGraph(golf, "Play", intNo = 10)
interactionGraph(golf, "Play", intNo = 10, speedUp = FALSE)
interactionGraph(golf, "Play", intNo = 10, speedUp = TRUE)

Calculates 3-Way Interactions

Description

Formula: I(X;Y;C) = I(X,Y;C) - IG(X;C) - IG(Y;C), where I(X;Y;C) is 3-way Interaction gain of the attributes X and Y, given the context (i.e. class) attribute C. Hence, I(X,Y;C) is a joint 2-way interaction gain (i.e. Information Gain) of the attributes X and Y, and I(X;C) and I(Y;C) are 2-way Interaction gains (i.e. Information Gains) of the attributes X and Y, respectively.

Usage

interactions3Way(df, classAtt, speedUp = FALSE)

Arguments

df

A discrete data.frame

classAtt

A class column of the df (string)

speedUp

A (boolean) parameter. If TRUE, indicates whether the pairs of attributes with Information Gain equal to zero (on the 4th decimal) should be pruned. This speeds up calculations for larger datasets. By default it is turned off (i.e. set to FALSE).

Value

A list with a: 1) data frame with 3-way interactions, 2)list of 2-way interactions of the input attributes

Examples

interactions3Way(golf, "Play")
interactions3Way(golf, "Play", speedUp = TRUE)
interactions3Way(golf, "Play", speedUp = FALSE)

Tests if data.frame is discrete (i.e. all of its columns are factors)

Description

Tests if data.frame is discrete (i.e. all of its columns are factors)

Usage

isDiscreteDataFrame(df)

Arguments

df

A data.frame

Value

Boolean: TRUE if all columns of the data.frame df are factors, FALSE otherwise; If the provided df object is of other type than data.frame, the function throws an error.

Examples

isDiscreteDataFrame(golf)

Plots Interaction graph

Description

Plots Interaction graph

Usage

plotIntGraph(ig)

Arguments

ig

Interaction graph

Value

Plots the ig

Examples

plotIntGraph(interactionGraph(golf, "Play", intNo = 10))

Print generic method for Interaction Graph (S3 class)

Description

Print generic method for Interaction Graph (S3 class)

Usage

## S3 method for class 'ig'
print(intGraph)

Arguments

intGraph

An (ig) object

Value

Print (ig) object

Print generic method for Interaction Graph Edges (S3 class)

Description

Print generic method for Interaction Graph Edges (S3 class)

Usage

## S3 method for class 'igEdge'
print(edge)

Arguments

edge

An (igEdge) object

Value

Print (igEdge) object

Print generic method for Interaction Graph Nodes (S3 class)

Description

Print generic method for Interaction Graph Nodes (S3 class)

Usage

## S3 method for class 'igNode'
print(node)

Arguments

node

An (igNode) object

Value

Print (igNode) object

toString() generic method for Interaction Graph (S3 class)

Description

toString() generic method for Interaction Graph (S3 class)

Usage

## S3 method for class 'ig'
toString(intGraph)

Arguments

intGraph

An ig object

Value

A character object made of the provided ig object

toString() generic method for Interaction Graph Edges (S3 class)

Description

toString() generic method for Interaction Graph Edges (S3 class)

Usage

## S3 method for class 'igEdge'
toString(edge)

Arguments

edge

An (igEdge) object

Value

(character) object made of the provided (igEdge) object

toString() generic method for Interaction Graph Nodes (S3 class)

Description

toString() generic method for Interaction Graph Nodes (S3 class)

Usage

## S3 method for class 'igNode'
toString(node)

Arguments

node

An (igNode) object

Value

(character) object made of the provided (igNode) object