baobab

An implementation of the ID3 (Iterative Dichotomiser 3) in Ruby

Use RubyGems:

gem install baobab

rake tests

Load baobab:

require `baobab` # if installed as a gem
load `baobab.rb` # if using a copy of the repo

Create your dataset from a JSON file. Your dataset should be a list of objects variable names and values as key-value pairs. See examples in the sample tests.

# A dataset on whether or not we go out to play depending on the outlook,
# humidity, and if it's windy
dataset = Dataset::from_json('test/weather.json')

Create a tree based on the dataaset and provide the name of the class variable.

tree = DecisionTree.new dataset, 'play'

Check your decision tree:

puts tree.to_s
# ROOT (0.94) # the number parenthesized is Shannon's entropy
#   outlook => rainy (0.694)
#     windy => TRUE (0.0)
#       play => no (0.0)
#     windy => FALSE (0.0)
#       play => yes (0.0)
#   outlook => overcast (0.694)
#     play => yes (0.0)
#   outlook => sunny (0.694)
#     humidity => normal (0.0)
#       play => yes (0.0)
#     humidity => high (0.0)
#       play => no (0.0)

You can make queries to your tree:

tree.query({"windy" => "yes", "outlook" => "rainy", "humidity" => "high"})
# no <-- we won't go out to play :(

Fun!

The weather dataset has been adapted from the weather.nominal.arff that comes shipped with Weka.

The transportation dataset was taken from the example data in https://www.youtube.com/watch?v=wL9aogTuZw8.

The breast cancer dataset is adapted from the breast-cancer.arff file that comes shipped with Weka (adapted so it doesn't have unknown values). It should be attributed to:

Matjaz Zwitter & Milan Soklic (physicians), Institute of Oncology, University Medical Center, Ljubljana, Yugoslavia. Donors: Ming Tan and Jeff Schlimmer (Jeffrey.Schlimmer@a.gp.cs.cmu.edu). Date: 11 July 1988.