[R-pkgs] Introducing The New Package 'forestRK'

[Hyunjin Cho]

Hello,


I am pleased to announce my new package forestRK. The package implements Forest-RK algorithm discussed in the paper "Forest-RK: A New Random Forest Induction Method" by Simon Bernard, Laurent Heutte, Sebastien Adam, 4th International Conference on Intelligent Computing (ICIC), Sep 2008, Shanghai, China, pp.430-437, to various datasets for classification.


Some of the forestRK functions were built based on the discussion:

  https://stats.stackexchange.com/questions/168964/building-a-regression-tree-with-r-from-scratch/168967#168967


Examples of functions included in the new forestRK package are (there are 17 functions in total in this package):


1. construct.treeRK: Builds a single decision tree after implementing the RK (random �K�) algorithm (i.e. builds �rktree�);


2. pred.treeRK: Makes predictions on the test observations based on the �rktree� model in question;


3. draw.treeRK: Makes igraph plot of a �rktree�.


4. forestRK: Builds a Forest-RK model;


5. pred.forestRK: Makes predictions on the test observations by using the Forest-RK algorithm;



6. mds.plot.forestRK: Generate 2D Multi-Dimensional Scaling plot of the test observations, where the test observations are colour coded by their predicted class type indicated in the pred.forestRK object;


7. importance.forestRK: Calculate Gini Importance of each covariate based on a forestRK model;


8. importance.plot.forestRK: Generate Importance ggplot of the covariates.


The forestRK package also provides tools to encode non-numeric dataset into a numeric one via Numeric Encoding or Binary Encoding.


For more information about the new forestRK package, please visit:


https://cran.r-project.org/web/packages/forestRK/index.html (CRAN)

https://github.com/h56cho/forestRK (Github)


Hyunjin Cho

	[[alternative HTML version deleted]]