def check_ID3():
attribute_metadata = [{'name': "winner",'is_nominal': True},{'name': "opprundifferential",'is_nominal': False}]
data_set = [[1, 0.27], [0, 0.42], [0, 0.86], [0, 0.68], [0, 0.04], [1, 0.01], [1, 0.33], [1, 0.42], [1, 0.42], [0, 0.51], [1, 0.4]]
numerical_splits_count = [5, 5]
n = ID3(data_set, attribute_metadata, numerical_splits_count, 0)
fails = 0;
if n and n.label == 1:
print "Passed 1"
else:
print "Failed 1"
fails += 1
attribute_metadata = [{'name': "winner",'is_nominal': True},{'name': "opprundifferential",'is_nominal': False}]
data_set = [[1, 0.27], [0, 0.42], [0, 0.86], [0, 0.68], [0, 0.04], [1, 0.01], [1, 0.33], [1, 0.42], [1, 0.42], [0, 0.51], [1, 0.4]]
numerical_splits_count = [1, 1]
n = ID3(data_set, attribute_metadata, numerical_splits_count, 5)
if n and [n.classify(x) == x[0] for x in data_set] == [True, False, True, True, False, True, True, True, True, True, True]:
print "Passed 2"
else:
print "Failed 2"
fails += 1
attribute_metadata = [{'name': "winner",'is_nominal': True},{'name': "opprundifferential",'is_nominal': False}]
data_set = [[1, 0.27], [0, 0.42], [0, 0.86], [0, 0.68], [0, 0.04], [1, 0.01], [1, 0.33], [1, 0.42], [1, 0.42], [0, 0.51], [1, 0.4]]
numerical_splits_count = [5, 5]
n = ID3(data_set, attribute_metadata, numerical_splits_count, 5)
if n and [n.classify(x) == x[0] for x in data_set] == [True, False, True, True, True, True, True, True, True, True, True]:
print "Passed 3"
else:
print "Failed 3"
fails += 1
if fails > 0:
print "not all tests passed, please see ID3."
else:
print "all tests passed."
def decision_tree_driver(train, validate = False, predict = False, prune = False,
limit_splits_on_numerical = False, limit_depth = False, print_tree = False,
print_dnf = False, learning_curve = False):
train_set, attribute_metadata = parse(train, False)
if limit_splits_on_numerical != False:
numerical_splits_count = [limit_splits_on_numerical] * len(attribute_metadata)
else:
numerical_splits_count = [float("inf")] * len(attribute_metadata)
if limit_depth != False:
depth = limit_depth
else:
depth = float("inf")
print "###\n# Training Tree\n###"
# call the ID3 classification algorithm with the appropriate options
tree = ID3(train_set, attribute_metadata, numerical_splits_count, depth)
print '\n'
# call reduced error pruning using the pruning set
if prune != False:
print '###\n# Pruning\n###'
pruning_set, _ = parse(prune, False)
n = Node()
reduced_error_pruning(tree,train_set,pruning_set, 0, n)
print ''
# print tree visually
if print_tree:
print '###\n# Decision Tree\n###'
cursor = open('./output/tree.txt','w+')
cursor.write(tree.print_tree())
cursor.close()
print 'Decision Tree written to /output/tree'
print ''
# print tree in disjunctive normalized form
if print_dnf:
print '###\n# Decision Tree as DNF\n###'
cursor = open('./output/DNF.txt','w+')
cursor.write(tree.print_dnf_tree())
cursor.close()
print 'Decision Tree written to /output/DNF'
print ''
# test tree accuracy on validation set
if validate != False:
print '###\n# Validating\n###'
validate_set, _ = parse(validate, False)
accuracy = validation_accuracy(tree,validate_set)
print "Accuracy on validation set: " + str(accuracy)
print ''
# generate predictions on the test set
if predict != False:
print '###\n# Generating Predictions on Test Set\n###'
create_predictions(tree, predict)
print ''
# generate a learning curve using the validation set
"""if learning_curve and validate:
def decision_tree_driver(train, validate = False, predict = False, prune = False,
limit_splits_on_numerical = False, limit_depth = False, print_tree = False,
print_dnf = False, learning_curve = False):
train_set, attribute_metadata = parse('D:/2016 Spring/349 Machine Learning/Problem Set 2/PS2.code/data/test_btrain.csv', False)
train_set = handle_missing_value(train_set,attribute_metadata)
if limit_splits_on_numerical != False:
numerical_splits_count = [limit_splits_on_numerical] * len(attribute_metadata)
else:
numerical_splits_count = [float("inf")] * len(attribute_metadata)
if limit_depth != False:
depth = limit_depth
else:
depth = float("inf")
print "###\n# Training Tree\n###"
# call the ID3 classification algorithm with the appropriate options
tree = ID3(train_set, attribute_metadata, numerical_splits_count, depth)
print '\n'
# call reduced error pruning using the pruning set
# if prune != False:
print '###\n# Pruning\n###'
pruning_set, _ = parse('D:/2016 Spring/349 Machine Learning/Problem Set 2/PS2.code/data/test_bvalidate.csv', False)
pruning_set = handle_missing_value(pruning_set,attribute_metadata)
accuracy = validation_accuracy(tree,pruning_set)
print(tree)
print "Accuracy on validation set of original tree: " + str(accuracy)
_ , newtree = reduced_error_pruning(tree,train_set,pruning_set)
print ''
# print tree visually
# if print_tree:
# print '###\n# Decision Tree\n###'
# cursor = open('./output/tree.txt','w+')
# cursor.write(tree.print_tree())
# cursor.close()
# print 'Decision Tree written to /output/tree'
# print ''
# print tree in disjunctive normalized form
# if print_dnf:
# print '###\n# Decision Tree as DNF\n###'
# cursor = open('./output/DNF.txt','w+')
# cursor.write(tree.print_dnf_tree())
# cursor.close()
# print 'Decision Tree written to /output/DNF'
# print ''
# test tree accuracy on validation set
# if validate != False:
print '###\n# Validating\n###'
accuracy2 = validation_accuracy(newtree,pruning_set)
print(newtree)
print "Accuracy on validation set of new tree: " + str(accuracy2)
print ''
def decision_tree_driver(train,
validate=False,
predict=False,
new=False,
prune=False,
limit_splits_on_numerical=False,
limit_depth=False,
print_tree=False,
print_dnf=False,
learning_curve=False):
train_set, attribute_metadata = parse(train, False)
if limit_splits_on_numerical != False:
numerical_splits_count = [limit_splits_on_numerical
] * len(attribute_metadata)
else:
numerical_splits_count = [float("inf")] * len(attribute_metadata)
if limit_depth != False:
depth = limit_depth
else:
depth = float("inf")
origin_splits_count = copy.deepcopy(numerical_splits_count)
print "###\n# Training Tree\n###"
# call the ID3 classification algorithm with the appropriate options
tree = ID3(train_set, attribute_metadata, numerical_splits_count, depth)
print 'finish'
if validate != False:
print '###\n# Validating\n###'
validate_set, _ = parse(validate, False)
accuracy = validation_accuracy(
tree, validate_set, attribute_metadata) #add attribute_metadata
print "Accuracy on validation set: " + str(accuracy)
print ''
# call reduced error pruning using the pruning set
if prune != False:
print '###\n# Pruning\n###'
pruning_set, _ = parse(prune, False)
temptree = copy.deepcopy(tree)
temp_origintree = temptree
origintree = tree
reduced_error_pruning(temptree, temp_origintree, tree, origintree,
train_set, pruning_set, attribute_metadata)
print ''
# print tree visually
if print_tree:
print '###\n# Decision Tree\n###'
cursor = open('./output/tree.txt', 'w+')
cursor.write(tree.print_tree())
cursor.close()
print 'Decision Tree written to /output/tree'
print ''
# print tree in disjunctive normalized form
if print_dnf:
print '###\n# Decision Tree as DNF\n###'
cursor = open('./output/DNF.txt', 'w+')
cursor.write(tree.print_dnf_tree())
cursor.close()
print 'Decision Tree written to /output/DNF'
print ''
# test tree accuracy on validation set
if validate != False:
print '###\n# Validating\n###'
validate_set, _ = parse(validate, False)
accuracy = validation_accuracy(
tree, validate_set, attribute_metadata) #add attribute_metadata
print "Accuracy on validation set: " + str(accuracy)
print ''
# generate predictions on the test set
# if predict != False:
# print '###\n# Generating Predictions on Test Set\n###'
# create_predictions(tree, predict, new) #add new
# print ''
# generate a learning curve using the validation set
if learning_curve and validate:
print '###\n# Generating Learning Curve\n###'
iterations = 3 # number of times to test each size
get_graph(train_set, attribute_metadata, validate_set,
origin_splits_count, depth, 3, 0,
learning_curve['upper_bound'], learning_curve['increment'])
print ''
def decision_tree_driver(train, validate = False, predict = False, prune = False,
limit_splits_on_numerical = False, limit_depth = False, print_tree = False,
print_dnf = False, learning_curve = False):
train_set, attribute_metadata = parse(train, False)
if limit_splits_on_numerical != False:
numerical_splits_count = [limit_splits_on_numerical] * len(attribute_metadata)
else:
numerical_splits_count = [float("inf")] * len(attribute_metadata)
if limit_depth != False:
depth = limit_depth
else:
depth = float("inf")
print "###\n# Training Tree\n###"
# call the ID3 classification algorithm with the appropriate options
tree = ID3(train_set, attribute_metadata, numerical_splits_count, depth)
print '\n'
print "Nodes before pruning: " + str(tree.num_nodes())
# # call reduced error pruning using the pruning set
if prune != False:
print '###\n# Pruning\n###'
pruning_set, _ = parse(prune, False)
reduced_error_pruning(tree,train_set,pruning_set)
print ''
print "Nodes after pruning: " + str(tree.num_nodes())
# print tree visually
if print_tree:
print '###\n# Decision Tree\n###'
cursor = open('./output/tree.txt','w+')
cursor.write(tree.print_tree())
cursor.close()
print 'Decision Tree written to /output/tree'
print ''
# print tree in disjunctive normalized form
if print_dnf:
print '###\n# Decision Tree as DNF\n###'
cursor = open('./output/DNF.txt','w+')
print tree.print_dnf_tree()
# cursor.write(final)
# cursor.close()
print 'Decision Tree written to /output/DNF'
print ''
# test tree accuracy on validation set
if validate != False:
print '###\n# Validating\n###'
train_set, _ = parse(train, False)
accuracy = validation_accuracy(tree, train_set)
print "Accuracy on training set: " + str(accuracy)
validate_set, _ = parse(validate, False)
accuracy = validation_accuracy(tree,validate_set)
print "Accuracy on validation set: " + str(accuracy)
print ''
# generate predictions on the test set
if predict != False:
print '###\n# Generating Predictions on Test Set\n###'
create_predictions(tree, predict)
print ''
# generate a learning curve using the validation set
if learning_curve and validate:
print '###\n# Generating Learning Curve\n###'
iterations = 2 # number of times to test each size
print get_graph_data(train_set, attribute_metadata, validate_set, numerical_splits_count, iterations, [0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0])
# get_graph(train_set, attribute_metadata, validate_set,
# numerical_splits_count, depth, 5, 0, learning_curve['upper_bound'],
# learning_curve['increment'])
print ''
from modules.predictions import * from modules.pickled import * from modules.parse import * from modules.node import * import matplotlib.pyplot as plt # Import training and validation datasets TRAINING_SET_PATH = 'data/btrain.csv' VALIDATION_SET_PATH = 'data/bvalidate.csv' TEST_SET_PATH = 'data/btest.csv' data, attr = parse(TRAINING_SET_PATH, True) validate_data, validate_attr = parse(VALIDATION_SET_PATH, True) # Train initial tree and print # Current best values for accuracy are below tree = ID3(data, attr, 14*[3], 5) print 'Question 2' print 'DNF form of initial tree trained on ' + TRAINING_SET_PATH + ':' tree.print_dnf_tree() print '\r\n' # Prune initial tree and print new tree pruned_tree = reduced_error_pruning(tree,data,validate_data) print 'Question 5' print 'DNF form of reduced-error pruned tree:' pruned_tree.print_dnf_tree() print '\r\n' # Calculate validation accuracy of initial tree and print print 'Question 7' print 'Initial tree validation accuracy:'
def decision_tree_driver(train, validate=False, predict=False, prune=False,
limit_splits_on_numerical=False, limit_depth=False, print_tree=False,
print_dnf=False, learning_curve=False):
train_set, attribute_metadata = parse(train, False)
if limit_splits_on_numerical != False:
numerical_splits_count = [limit_splits_on_numerical] * len(attribute_metadata)
else:
numerical_splits_count = [float("inf")] * len(attribute_metadata)
if limit_depth != False:
depth = limit_depth
else:
depth = float("inf")
print "###\n# Training Tree\n###"
# call the ID3 classification algorithm with the appropriate options
tree = ID3(train_set, attribute_metadata, numerical_splits_count, depth)
print '\n'
# print tree visually
if print_tree:
print '###\n# Decision Tree\n###'
cursor = open('./output/tree.txt', 'w+')
cursor.write(tree.print_tree())
cursor.close()
print 'Decision Tree written to /output/tree'
print ''
# print tree in disjunctive normalized form
if print_dnf:
print '###\n# Decision Tree as DNF\n###'
cursor = open('./output/DNF.txt', 'w+')
cursor.write(tree.print_dnf_tree())
cursor.close()
print 'Decision Tree written to /output/DNF'
print ''
# test tree accuracy on validation set
if validate != False:
print '###\n# Validating\n###'
validate_set, _ = parse(validate, False)
accuracy = validation_accuracy(tree, validate_set)
print "Accuracy on training set: " + str(validation_accuracy(tree, train_set))
print "Accuracy on validation set: " + str(accuracy)
print ''
# call reduced error pruning using the pruning set
if prune != False:
print '###\n# Pruning\n###'
pruning_set, _ = parse(prune, False)
reduced_error_pruning(tree, pruning_set)
print ''
# print tree visually
if print_tree:
print '###\n# Decision Tree\n###'
cursor = open('./output/prune_tree.txt', 'w+')
cursor.write(tree.print_tree())
cursor.close()
print 'Decision Tree written to /output/prune_tree'
print ''
# print tree in disjunctive normalized form
if print_dnf:
print '###\n# Decision Tree as DNF\n###'
cursor = open('./output/prune_DNF.txt', 'w+')
cursor.write(tree.print_dnf_tree())
cursor.close()
print 'Decision Tree written to /output/prune_DNF'
print ''
# test tree accuracy on validation set
if validate != False:
print '###\n# Validating\n###'
validate_set, _ = parse(validate, False)
accuracy = validation_accuracy(tree, validate_set)
print "Accuracy on training set: " + str(validation_accuracy(tree, train_set))
print "Accuracy on validation set: " + str(accuracy)
print ''
# generate predictions on the test set
if predict != False:
print '###\n# Generating Predictions on Test Set\n###'
with open('./output/predictions.csv', 'w+') as cursor:
writer = csv.writer(cursor)
fieldnames = ['winner', ' winpercent', ' oppwinpercent', ' weather', ' temperature', ' numinjured', ' oppnuminjured',
' startingpitcher', ' oppstartingpitcher', ' dayssincegame', ' oppdayssincegame', ' homeaway',
' rundifferential', ' opprundifferential']
writer.writerow(fieldnames)
writer.writerows(create_predictions(tree, predict))
print ''
# generate a learning curve using the validation set
if learning_curve and validate:
print '###\n# Generating Learning Curve\n###'
validate_set, _ = parse(validate, False)
iterations = 5 # number of times to test each size
get_graph(train_set, attribute_metadata, validate_set,
numerical_splits_count, depth, iterations, 0, learning_curve['upper_bound'],
learning_curve['increment'])
print ''
