## Contents of the ZIP file

  The files in the ZIP file, will be extracted into a directory
  containing  all code and data that is necessary to replicate all
  experiments reported in the paper "Re-sampling Strategies for Regression".

  The folder contains  the following files:

  - allDataSets.Rdata        the used data sets 

  - cvAuxs.R                 the auxiliary functions used by the different learning routines to get the statistics estimated by CV

  - extremesResampling.R     the smote and under-sampling strategies 

  - newCVexps.R              the script for running a set of experiences

  - PaperGraphs.R 	     to replicate the graphs in the paper

## Necessary software

  To replicate these experiments you will need a working installation
  of R (check www.r-project.org if you need to download and install it)

  In your R installation you also need to install the following
  additional R packages:
  - DMwR
  - e1071
  - earth
  - randomForest
  - uba

  All the above packages except "uba", can be installed from CRAN
  Repository directly as ay "normal" R package. Essentially you issue
  the command :

> install.packages(c("DMwR", "e1071", "earth", "randomForest"))

  within R.

  The package uba needs to be installed from a tar.gz file that you
  can download from http://www.dcc.fc.up.pt/~rpribeiro/uba/.
  Download the tar.gz file into your folder and then issue:

> install.packages("uba_0.7.5.tar.gz",repos=NULL,dependencies=T)

  To replicate the graphs in the paper you also need to install the
  packages:

  - plyr
  - ggplot2


## Running the experiences:
  
  To run all experiments you execute R in the folder with the data and
  code and then issue the command:

> source("newCVexps.R")
  
  Alternatively, you may run the experiments directly from a Linux terminal
  (useful if you want to logout because the experiments take a long
  time to run):

$ > nohup R --vanilla --quiet < newCVexps.R &


##  Running a subset of the experiences:
  
  For running only a subset of the experiences the newCVexps.R must be altered.
  The VARS parameters can be altered to incorporate other parameters for the svm, randomForest or earth;
  the TODO variable can be altered for containing only a subset of the learning systems;
  the todoDSs variable can also be altered to contain just a subset of the data sets;
  finally, the parameters used in smote and undersampling can also be altered in the variants 
  defined on the main loop by removing, adding or changing the values.
  After performing the desired changes, run the experiences as described previously.