A C4.5 Decision Tree python implementation with validation, pruning, and attribute multi-splitting Contributors: Ryan Madden and Ally Cody

python 2.7.6 Download

• btrain.csv, bvalidate.csv, btest.csv - The training, validation, and testing sets used for building and testing the program
• decision-tree.py - The decision tree program
• datatypes.csv - A metadata file that indicates (with comma separated true/false entries) which attributes are numeric (true) and nominal (false) Note: You must edit this file or supply your own if using a different dataset than the one provided

decision-tree.py accepts parameters passed via the command line. The possible paramters are:

• Filename for training (Required, must be the first argument after 'python decision-tree.py')
• Classifier name (Optional, by default the classifier is the last column of the dataset)
• Datatype flag (-d) followed by datatype filename (Optional, defaults to 'datatypes.csv')
• Print flag (-s) (Optional, causes the dataset)
• Validate flag (-v) followed by validate filename (Optional, specifies file to use for validation)
• Test flag (-t) followed by test filename (Optional, specifies file to use for testing)
• Pruning flag (-p) (Optional, you must include a validation file in order to prune)

#####Example 1

python decision-tree.py btrain.csv -v bvalidate.csv -p -t btest.csv

This command runs decision-tree.py with btrain.csv as the training set, bvalidate.csv as the validation set, btest.csv as the test set, and pruning enabled. The classifier is not specified so it defaults to the last column in the training set. No datatypes file is specified so it defaults to datatypes.csv. Printing is not enabled. #####Example 2

python decision-tree.py btrain.csv winner -d datatypes.csv -v bvalidate.csv -p -s

This command runs decision-tree.py with btrain.csv as the training set, 'winner' as the classifier, datatypes.csv as the datatypes file, bvalidate.csv as the validation file, pruning enabled, and printing enabled. Testing is not enabled for this run