decision_tree_example

This code snippets are an example on how to build a decision tree using Shannon Entropy.

Files

• index.js: contains an example on how to use the library
• library/best_feature_selector.js : Method to select the best feature in a data set based on the information gain calculated thanks to the Shannon Entropy
• library/data_set_splitter.js : Method to split a dataset based on one feature and one value of this feature
• library/shannon_entropy.js : Method to compute shannon entropy on a dataset.
• library/tree_generator.js : The core module that contains an iterative function that will generate the tree

How to test it ?

$ git clone [email protected]:maximilienandile/decision_tree_example.git
$ cd decision_tree_example
$ node index.js

Convention for the dataset

The dataset should be an array. This array contains one object per training example. Each object should have a property called label that represents the class of the training example.

var training_data = [
  {"has_a_car":"yes", "has_a_house":"yes","label":"good_borrower"},
  {"has_a_car":"yes", "has_a_house":"yes","label":"good_borrower"},
  {"has_a_car":"yes", "has_a_house":"no","label":"bad_borrower"},
  {"has_a_car":"no", "has_a_house":"yes","label":"bad_borrower"},
  {"has_a_car":"no", "has_a_house":"yes","label":"bad_borrower"}
]

Here the features are : has_a_car and has_a_house.

The classes are good_borrower or bad_borrower

TO DO

• Generate a visual representation of the tree
• Improve the algorithm
• Try other algorithms
• Try other methods to compute information gain