cyrusaf/decision_tree

Decision Tree Starter Code

#Credits: Philip Lan, Christopher Li, Alex Wang, Jacob Samson, Nishant Subramani

Building

To build the project, you must pass it a dictionary of inputs.Below are listed the options available to be passed.

Documentation

train Required. Should be passed a path from the decision_tree_driver.py file to the file to be trained on. validate Should be passed a path from the decision_tree_driver.py file to the file to be validated on or left blank. predict Should be passed a path from the decision_tree_driver.py file to the file to predict outputs. The generated output will be created and named after this file in the format of file_predictions.csv or left blank. limit_splits_on_numerical Should be passed an integer that limits the number of splits that can be made on each numerical attribute. limit_depth Should be passed an integer that limits depth of the decision tree. print_tree Should be passed a boolean which will print a tree in readable format if True. print_dnf Should be passed a boolean which will print a disjunctive normalized form of the decision tree if True. prune Should be passed a boolean which prune the tree with the reduced error pruning method if True. Requires validate. learning_curve Should be passed a dictionary of upper bound and increment or left blank. This will take a portion of the training data and validates it on the validation set and will be done over a range of partials. The plot will be saved in the output directory.

Example

Below are the commands necessary to run the program. The options dictionary is customizable based on the documentation above.

execfile("decision_tree_driver.py")
options = {
    'train' : 'data/btrain.csv',
    'validate': 'data/bvalidate.csv',
    'predict': 'data/btest.csv',
    'limit_splits_on_numerical': 5,
    'limit_depth': 20,
    'print_tree': True,
    'print_dnf' : True,
    'prune' : True,
    'learning_curve' : {
    	'upper_bound' : 0.01,
    	'increment' : 0.001
	}
}
decision_tree_driver(options)

The only variable required is train, so the program will not output anything, it can also execute with only the train key:

execfile("decision_tree_driver.py")
options = {
	'train': 'data/btrain.csv'
}
decision_tree_driver(options)