Decision Tree Trainer

Alberto Serrano

Description

A simple decision tree trainer that assumes the data being fed is a set of labeled muffin and cupcake recipes. Uses the Gini Index impurity function to make the "best" split. Generates a second file that is able classify the type of recipe.

Stopping Criteria

There are two stopping criterias used in the implementation of the decision tree trainer:

• If the type of recipes left in the data are more than 90% muffin or cupcake, then stop expanding and add a leaf node.
• If there are less than 3 recipes left, then stop expanding and insert a leaf node.

Future work

• Generalize algorithm to accept user defined classes.
• Make trainer able to create n-class classifier.
• Use other impurity functions in place of Gini Index.