Rfviz: An Interactive Visualization Package for Random Forests in R

Description

Rfviz is an interactive package and toolkit in R, using TclTK code on the backend, to help in viewing and interpreting the results Random Forests for both Supervised Classification and Regression in a user-friendly way.

Details

Currently, rfviz implements the following statistical graphs, with functions to view any combination of the plots:

The three plots are:

1. The classic multidimensionally scaled proximities are plotted as a 3-D XYZ scatterplot.

2. The raw input data is plotted in a parallel coordinate plot.

3. The local importance scores of each observation are plotted in a parallel coordinate plot.

rfviz is built using the package Loon on the backend, and implements the random forests algorithm.

For detailed instructions in the use of these plots in this package, visit https://github.com/chriskuchar/rfviz/blob/master/Rfviz.md

Note

For instructions on how to use randomForests, use ?randomForest. For more information on loon, use ?loon.

Author(s)

Chris Kuchar chrisjkuchar@gmail.com, based on original Java graphics by Leo Breiman and Adele Cutler.

References

Liaw A, Wiener M (2002). “Classification and Regression by randomForest.” _R News_, *2*(3), 18-22. https://CRAN.R-project.org/doc/Rnews/

Waddell A, Oldford R. Wayne (2018). "loon: Interactive Statistical Data Visualization" https://github.com/waddella/loon

Breiman, L. (2001), Random Forests, Machine Learning 45(1), 5-32.

Breiman, L (2002), “Manual On Setting Up, Using, And Understanding Random Forests V3.1”, https://www.stat.berkeley.edu/~breiman/Using_random_forests_V3.1.pdf

Breiman, L., Cutler, A., Random Forests Graphics. https://www.stat.berkeley.edu/~breiman/RandomForests/cc_graphics.htm

See Also

randomForest, rf_prep, rf_viz, l_plot3D, l_serialaxes

Glass Identification Data Set

Description

A dataset containing 6 types of glass; defined in terms of their oxide content

Usage

glass

Format

A data frame with 214 rows and 10 variables:

RI

Refractive Index

Na

Sodium (unit measurement: weight percent in corresponding oxide, as are attributes 4-10)

Mg

Magnesium

Al

Aluminum

Si

Silicon

K

Potassium

Ca

Calcium

Ba

Barium

Fe

Iron

Type

Class Attribute

Source

https://archive.ics.uci.edu/ml/datasets/glass+identification

A function to create Random Forests output in preparation for visualization with rf_viz

Description

A function using Random Forests which outputs a list of the Random Forests output, the predictor variables data, and response variable data.

Usage

rf_prep(x, y = NULL, ...)

Arguments

Value

The parallel coordinate plots of the input data, the local importance scores, and the 3-D XYZ classic multidimensional scaling proximities from the output of the random forest algorithm.

Note

For instructions on how to use randomForests, use ?randomForest. For more information on loon, use ?loon.

For detailed instructions in the use of these plots in this package, visit https://github.com/chriskuchar/rfviz/blob/master/Rfviz.md

Author(s)

Chris Kuchar chrisjkuchar@gmail.com, based on original Java graphics by Leo Breiman and Adele Cutler.

References

Liaw A, Wiener M (2002). “Classification and Regression by randomForest.” _R News_, *2*(3), 18-22. https://CRAN.R-project.org/doc/Rnews/

Waddell A, Oldford R. Wayne (2018). "loon: Interactive Statistical Data Visualization" https://github.com/waddella/loon

Breiman, L. (2001), Random Forests, Machine Learning 45(1), 5-32.

Breiman, L (2002), “Manual On Setting Up, Using, And Understanding Random Forests V3.1”, https://www.stat.berkeley.edu/~breiman/Using_random_forests_V3.1.pdf

Breiman, L., Cutler, A., Random Forests Graphics. https://www.stat.berkeley.edu/~breiman/RandomForests/cc_graphics.htm

See Also

randomForest, rf_viz, l_plot3D, l_serialaxes

Examples

#Preparation for classification with Iris data set
rfprep <- rf_prep(x=iris[,1:4], y=iris$Species)

#Preparation for regression with mtcars data set
rfprep <- rf_prep(x=mtcars[,-1], y=mtcars$mpg)

#Preparation for the unsupervised case with Iris data set
rfprep <- rf_prep(x=iris[,1:4], y=NULL)

Random Forest Plots for interpreting Random Forests output

Description

The Input Data, Local Importance Scores, and Classic Multidimensional Scaling Plots

Usage

rf_viz(rfprep, input = TRUE, imp = TRUE, cmd = TRUE, hl_color = "orange")

Arguments

Value

Any combination of the parallel coordinate plots of the input data, the local importance scores, and the 3-D XYZ classic multidimensional scaling proximities from the output of the random forest algorithm.

Note

For instructions on how to use randomForests, use ?randomForest. For more information on loon, use ?loon.

For detailed instructions in the use of these plots in this package, visit https://github.com/chriskuchar/rfviz/blob/master/Rfviz.md

Author(s)

Chris Kuchar chrisjkuchar@gmail.com, based on original Java graphics by Leo Breiman and Adele Cutler.

References

Liaw A, Wiener M (2002). “Classification and Regression by randomForest.” _R News_, *2*(3), 18-22. https://CRAN.R-project.org/doc/Rnews/

Waddell A, Oldford R. Wayne (2018). "loon: Interactive Statistical Data Visualization" https://github.com/waddella/loon

Breiman, L. (2001), Random Forests, Machine Learning 45(1), 5-32.

Breiman, L (2002), “Manual On Setting Up, Using, And Understanding Random Forests V3.1”, https://www.stat.berkeley.edu/~breiman/Using_random_forests_V3.1.pdf

Breiman, L., Cutler, A., Random Forests Graphics. https://www.stat.berkeley.edu/~breiman/RandomForests/cc_graphics.htm

See Also

randomForest, rf_prep, l_plot3D, l_serialaxes

Examples

#Classification with iris data set
rfprep <- rf_prep(x = iris[,1:4], y = iris$Species)

#View all three plots
Myrfplots <- rf_viz(rfprep, input = TRUE, imp = TRUE, cmd = TRUE, hl_color = 'orange')

#Select data on any of the plots then run:
iris[Myrfplots$input['selected'], ]
iris[Myrfplots$imp['selected'], ]
iris[Myrfplots$cmd['selected'], ]

#Rotate 3-D XYZ Scatterplot
#1. Click on 3-D XYZ Scatterplot
#2. Press 'r' on keyboard to enter rotation mode
#3. Click and drag mouse to rotate plot
#4. Press 'r' to leave rotation mode

#View only the Input Data and CMD Scaling Proximities Plots
Myrfplots <- rf_viz(rfprep, input = TRUE, imp = FALSE, cmd = TRUE, hl_color = 'orange')

#Regression with mtcars data set
rfprep2 <- rf_prep(x = mtcars[,-1], y = mtcars$mpg)

#View all three plots
Myrfplots <- rf_viz(rfprep2, input = TRUE, imp = TRUE, cmd = TRUE, hl_color = 'orange')

#Unsupervised clustering with iris data set 
rfprep <- rf_prep(x = iris[,1:4], y = NULL)

#View the Input Data and CMD Scaling Proximities Plots for the unsupervised case. 
#(Importance Scores Plot not valid here)
Myrfplots <- rf_viz(rfprep, input = TRUE, imp = FALSE, cmd = TRUE, hl_color = 'orange')