#### Supplemental R-Code for
####
#### (Psycho-)Analysis of Benchmark Experiments: A Formal Framework
#### for Investigating the Relationship between Data Sets and
#### Learning Algorithms
####
#### by Manuel J. A. Eugster, Friedrich Leisch and Carolin Strobl.

library(benchmark)
library(classInt)



### Load uci13 benchmark data:

load('uci13.RData')



### Categorize continuous variables:

categorize <- function(x) {
  y <- cut(x, breaks = classIntervals(x)$brks)
  levels(y) <- c('NA', levels(y))
  y[is.na(y)] <- 'NA'
  y
}

uci13.cat <- within(uci13, {
  input.n <- ordered(input.n)
  input.attr <- ordered(input.attr)
  input.factor.attr <- ordered(input.factor.attr)
  input.factor...bin <- ordered(input.factor...bin)
  input.numeric.attr <- ordered(input.numeric.attr)
  response.factor...cl <- ordered(response.factor...cl)

  input.factor...entropy <- categorize(input.factor...entropy)
  input.numeric.mac <- categorize(input.numeric.mac)
  input.numeric...skewness <- categorize(input.numeric...skewness)
  input.numeric...kurtosis <- categorize(input.numeric...kurtosis)
  response.factor...entropy <- categorize(response.factor...entropy)
  input2response.numeric2factor.fcc <- categorize(input2response.numeric2factor.fcc)
  input2response.numeric2factor.frac1 <- factor(input2response.numeric2factor.frac1)
  input2response.factor2factor.mi <- categorize(input2response.factor2factor.mi)
  input2response.factor2factor.enattr <- categorize(input2response.factor2factor.enattr)
  input2response.factor2factor.nsratio <- categorize(input2response.factor2factor.nsratio)
})



### Calculate tree:

tree <- bttree(preference ~ ., data = uci13.cat, verbose = TRUE, minsplit = 200)

load('uci13-tree.RData')  # precalculated tree