Add `nbins` argument to `information.gain()`

[pat-s]

Making it possible for the user to choose the bin size.

The bin size could optionally be tuned when information.gain() is used with other packages, e.g. mlr.

Other changes

• Added myself as a ctb
• Updated the URL field so that usethis::browse_github() works. Opinionated, not sure if you still want to link to your pkgdown site

Example

library(mlr)
#> Loading required package: ParamHelpers
library(FSelectorRcpp)

# generate data
task = dropFeatures(bh.task, "chas")

data = getTaskData(task)
x = data[getTaskFeatureNames(task)]
y = data[[getTaskTargetNames(task)]]

# equal = FALSE (default)
information_gain(x = x, y = y)
#> Warning in .information_gain.data.frame(x = x, y = y, type = type, equal
#> = equal, : Dependent variable is a numeric! It will be converted to
#> factor with simple factor(y). We do not discretize dependent variable
#> in FSelectorRcpp by default! You can choose equal frequency binning
#> discretization by setting equal argument to TRUE.
#>    attributes importance
#> 1        crim          0
#> 2          zn          0
#> 3       indus          0
#> 4         nox          0
#> 5          rm          0
#> 6         age          0
#> 7         dis          0
#> 8         rad          0
#> 9         tax          0
#> 10    ptratio          0
#> 11          b          0
#> 12      lstat          0

# equal = TRUE
information_gain(x = x, y = y, equal = TRUE)
#>    attributes importance
#> 1        crim  0.2433297
#> 2          zn  0.1259694
#> 3       indus  0.3493898
#> 4         nox  0.3575879
#> 5          rm  0.4123003
#> 6         age  0.2516128
#> 7         dis  0.1499913
#> 8         rad  0.1491553
#> 9         tax  0.2543226
#> 10    ptratio  0.2924772
#> 11          b  0.1011264
#> 12      lstat  0.5993705

# equal = TRUE, nbins = 10
information_gain(x = x, y = y, equal = TRUE, nbins = 10)
#>    attributes importance
#> 1        crim  0.2990401
#> 2          zn  0.1308077
#> 3       indus  0.3393358
#> 4         nox  0.3566715
#> 5          rm  0.3767511
#> 6         age  0.2052093
#> 7         dis  0.1916598
#> 8         rad  0.1700187
#> 9         tax  0.1988506
#> 10    ptratio  0.2356597
#> 11          b  0.1081561
#> 12      lstat  0.5917914

# equal = TRUE, nbins = 30
information_gain(x = x, y = y, equal = TRUE, nbins = 30)
#>    attributes importance
#> 1        crim  0.2534268
#> 2          zn  0.0000000
#> 3       indus  0.2637486
#> 4         nox  0.2436495
#> 5          rm  0.4489540
#> 6         age  0.2348189
#> 7         dis  0.2423434
#> 8         rad  0.0000000
#> 9         tax  0.2316402
#> 10    ptratio  0.2622159
#> 11          b  0.0000000
#> 12      lstat  0.6510120

^{Created on 2019-06-25 by the reprex package (v0.3.0)}

[MarcinKosinski]

CC @zzawadz

[MarcinKosinski]

Hey @pat-s I see an error on travis, however I see that's the dependencies issues. Let me restart the build and remove unneeded dependencies if that might help

[MarcinKosinski]

Thanks @pat-s for the update