def test_id3_ut24(self, tn=24):
"""
Tests the DT on play_tennis_data.csv and play_tennis_target.txt.
"""
print('\n============ ID3 UT {} ======================='.format(tn))
examples, colnames = bin_id3.parse_csv_file_into_examples('play_tennis.csv')
attribs = set(colnames[1:])
avt = bin_id3.construct_attrib_values_from_examples(examples, colnames[1:])
target_attrib = 'PlayTennis'
dtr = bin_id3.fit(examples, target_attrib, attribs, avt, True)
bin_id3.display_id3_node(dtr, '')
uexamples, ucolnames = bin_id3.parse_csv_file_into_examples('play_tennis_data.csv')
gt = []
with open('play_tennis_target.txt', 'r') as inf:
for ln in inf:
gt.append(ln.strip())
assert len(uexamples) == len(gt)
accurate_count = 0
for uex, gtt in zip(uexamples, gt):
pred = bin_id3.predict(dtr, uex)
if pred == gtt:
accurate_count += 1
accuracy = accurate_count/len(uexamples)
assert accuracy == 1.0
print('classification accuracy = {}'.format(accurate_count/len(uexamples)))
def test_id3_ut23(self, tn=23):
print('\n============ ID3 UT {} ======================='.format(tn))
global PLUS
global MINUS
### Let's learn a DT.
examples, colnames = bin_id3.parse_csv_file_into_examples('play_tennis.csv')
attribs = set(colnames[1:])
avt = bin_id3.construct_attrib_values_from_examples(examples, colnames[1:])
target_attrib = 'PlayTennis'
dtr = bin_id3.fit(examples, target_attrib, attribs, avt, True)
bin_id3.display_id3_node(dtr, '')
### Let's load unlabeled examples
unlbl_examples, unlbl_colnames = bin_id3.parse_csv_file_into_examples('play_tennis_unlbl.csv')
print(len(unlbl_examples))
print('Unlabeled Examples:\n')
for i, ex in enumerate(unlbl_examples):
print('{}) {}'.format(i+1, ex))
print('\nColumn names:')
for i, cn in enumerate(unlbl_colnames):
print('{}) {}'.format(i+1, cn))
prediction = bin_id3.predict(dtr, unlbl_examples[0])
assert prediction == MINUS
prediction = bin_id3.predict(dtr, unlbl_examples[1])
assert prediction == MINUS
prediction = bin_id3.predict(dtr, unlbl_examples[2])
assert prediction == PLUS
prediction = bin_id3.predict(dtr, unlbl_examples[3])
assert prediction == PLUS
prediction = bin_id3.predict(dtr, unlbl_examples[4])
assert prediction == PLUS
prediction = bin_id3.predict(dtr, unlbl_examples[5])
assert prediction == MINUS
prediction = bin_id3.predict(dtr, unlbl_examples[6])
assert prediction == PLUS
prediction = bin_id3.predict(dtr, unlbl_examples[7])
assert prediction == MINUS
prediction = bin_id3.predict(dtr, unlbl_examples[8])
assert prediction == PLUS
prediction = bin_id3.predict(dtr, unlbl_examples[9])
assert prediction == PLUS
prediction = bin_id3.predict(dtr, unlbl_examples[10])
assert prediction == PLUS
prediction = bin_id3.predict(dtr, unlbl_examples[11])
assert prediction == PLUS
prediction = bin_id3.predict(dtr, unlbl_examples[12])
assert prediction == PLUS
prediction = bin_id3.predict(dtr, unlbl_examples[13])
assert prediction == MINUS
print('\n============ ID3 UT {} passed ================'.format(tn))
def test_id3_ut20(self, tn=20):
"""
Let S =
1) {'Day': 'D1', 'Outlook': 'Sunny', 'Temperature': 'Hot', 'Humidity': 'High', 'Wind': 'Weak', 'PlayTennis': 'No'}
2) {'Day': 'D2', 'Outlook': 'Sunny', 'Temperature': 'Hot', 'Humidity': 'High', 'Wind': 'Strong', 'PlayTennis': 'No'}
3) {'Day': 'D8', 'Outlook': 'Sunny', 'Temperature': 'Mild', 'Humidity': 'High', 'Wind': 'Weak', 'PlayTennis': 'No'}
4) {'Day': 'D9', 'Outlook': 'Sunny', 'Temperature': 'Cool', 'Humidity': 'Normal', 'Wind': 'Weak', 'PlayTennis': 'Yes'}
5) {'Day': 'D11', 'Outlook': 'Sunny', 'Temperature': 'Mild', 'Humidity': 'Normal', 'Wind': 'Strong', 'PlayTennis': 'Yes'}
Find most common value for attribute PlayTennis
Find most common value for attribute Humdity
Find most common value for attribute Wind
Find most common value for attribute Temperature
"""
print('\n============ ID3 UT {} ======================='.format(tn))
examples, colnames = bin_id3.parse_csv_file_into_examples('play_tennis.csv')
outlook_sunny_examples = bin_id3.find_examples_given_attrib_val(examples, 'Outlook', 'Sunny')
avt = bin_id3.construct_attrib_values_from_examples(examples, colnames[1:])
atv, cnt = bin_id3.find_most_common_attrib_val(outlook_sunny_examples, 'PlayTennis', avt)
assert atv == 'No'
assert cnt == 3
atv, cnt = bin_id3.find_most_common_attrib_val(outlook_sunny_examples, 'Humidity', avt)
assert atv == 'High'
assert cnt == 3
atv, cnt = bin_id3.find_most_common_attrib_val(outlook_sunny_examples, 'Wind', avt)
assert atv == 'Weak'
assert cnt == 3
atv, cnt = bin_id3.find_most_common_attrib_val(outlook_sunny_examples, 'Temperature', avt)
assert atv == 'Hot' or atv == 'Mild'
assert cnt == 2
print('\n============ ID3 UT {} passed ================'.format(tn))
def test_id3_ut01(self, tn=1):
"""
Tests bin_id3.construct_attrib_values_from_examples().
"""
print('\n============ ID3 UT {} ======================='.format(tn))
examples, colnames = bin_id3.parse_csv_file_into_examples('play_tennis.csv')
avt = bin_id3.construct_attrib_values_from_examples(examples, colnames[1:])
print('Attribute --> Values:\n')
for k, v in avt.items():
print('{} --> {}'.format(k, v))
print('\n============ ID3 UT {} passed ================'.format(tn))
def test_id3_ut22(self, tn=22):
"""
Learn a decision tree with Binary ID3 and display it.
"""
print('\n============ ID3 UT {} ======================='.format(tn))
examples, colnames = bin_id3.parse_csv_file_into_examples('play_tennis.csv')
attribs = set(colnames[1:])
avt = bin_id3.construct_attrib_values_from_examples(examples, colnames[1:])
target_attrib = 'PlayTennis'
root = bin_id3.fit(examples, target_attrib, attribs, avt, True)
bin_id3.display_id3_node(root, '')
print('\n============ ID3 UT {} passed ================'.format(tn))
def test_id3_ut10(self, tn=10):
"""
Tests information gain of Temperature in the examples and target_attribute PlayTennis.
"""
print('\n============ ID3 UT {} ======================='.format(tn))
examples, colnames = bin_id3.parse_csv_file_into_examples('play_tennis.csv')
avt = bin_id3.construct_attrib_values_from_examples(examples, colnames[1:])
g = bin_id3.gain(examples, 'PlayTennis', 'Temperature', avt)
gt = 0.029222565658954647
err = 0.0001
print('gain(Temperature)={}'.format(g))
assert abs(g - gt) <= err
print('\n============ ID3 UT {} passed ================'.format(tn))
def test_id3_ut09(self, tn=9):
"""
Tests information gain of Outlook in the examples and target_attribute PlayTennis.
"""
print('\n============ ID3 UT {} ======================='.format(tn))
examples, colnames = bin_id3.parse_csv_file_into_examples('play_tennis.csv')
avt = bin_id3.construct_attrib_values_from_examples(examples, colnames[1:])
g = bin_id3.gain(examples, 'PlayTennis', 'Outlook', avt)
gt = 0.2467498197744391
err = 0.0001
print('gain(Outlook)={}'.format(g))
assert abs(g - gt) <= err
print('\n============ ID3 UT {} passed ================'.format(tn))
def test_id3_ut08(self, tn=8):
"""
Tests information gain of Humidity in the examples and target_attribute PlayTennis.
"""
print('\n============ ID3 UT {} ======================='.format(tn))
examples, colnames = bin_id3.parse_csv_file_into_examples('play_tennis.csv')
avt = bin_id3.construct_attrib_values_from_examples(examples, colnames[1:])
g = bin_id3.gain(examples, 'PlayTennis', 'Humidity', avt)
gt = 0.15183550136234136
err = 0.0001
print('gain(Humidity)={}'.format(g))
assert abs(g - gt) <= err
print('\n============ ID3 UT {} passed ================'.format(tn))
def test_id3_ut18(self, tn=18):
"""
Find all examples with Outlook=Rain and check that they include 3 positive and 2 negative
examples.
"""
print('\n============ ID3 UT {} ======================='.format(tn))
examples, colnames = bin_id3.parse_csv_file_into_examples('play_tennis.csv')
outlook_rain_examples = bin_id3.find_examples_given_attrib_val(examples, 'Outlook', 'Rain')
print('Examples with Outlook=Rain:\n')
for i, ex in enumerate(outlook_rain_examples):
print('{}) {}'.format(i+1, ex))
assert len(bin_id3.find_examples_given_attrib_val(outlook_rain_examples, 'PlayTennis', 'Yes')) == 3
assert len(bin_id3.find_examples_given_attrib_val(outlook_rain_examples, 'PlayTennis', 'No')) == 2
print('\n============ ID3 UT {} passed ================'.format(tn))
def test_id3_ut06a(self, tn=6):
"""
Finds entropy of the original examples with respect to target_attribute PlayTennis.
"""
print('\n============ ID3 UT {}a ======================='.format(tn))
examples, colnames = bin_id3.parse_csv_file_into_examples('play_tennis.csv')
avt = bin_id3.construct_attrib_values_from_examples(examples, colnames[1:])
H = bin_id3.entropy(examples, 'PlayTennis', avt)
## ground truth value for Entropy
gH = 0.9402859586706309
err = 0.0001
print('Entropy(S)={}'.format(H))
assert abs(H - gH) <= err
print('\n============ ID3 UT {}a passed ================'.format(tn))
def test_id3_ut14(self, tn=14):
"""
Find all examples with Outlook=Overcast and check that those examples consist of
4 postive examples and 0 negative examples.
"""
print('\n============ ID3 UT {} ======================='.format(tn))
examples, colnames = bin_id3.parse_csv_file_into_examples('play_tennis.csv')
outlook_over_examples = bin_id3.find_examples_given_attrib_val(examples, 'Outlook', 'Overcast')
print('Examples with Outlook=Overcast:\n')
for i, ex in enumerate(outlook_over_examples):
print('{}) {}'.format(i+1, ex))
assert len(bin_id3.find_examples_given_attrib_val(outlook_over_examples, 'PlayTennis', 'Yes')) == 4
assert len(bin_id3.find_examples_given_attrib_val(outlook_over_examples, 'PlayTennis', 'No')) == 0
print('\n============ ID3 UT {} passed ================'.format(tn))
def test_id3_ut12(self, tn=12):
"""
Find all examples with Outlook=Sunny and compute info gain of Temperature.
"""
print('\n============ ID3 UT {} ======================='.format(tn))
examples, colnames = bin_id3.parse_csv_file_into_examples('play_tennis.csv')
outlook_sunny_examples = bin_id3.find_examples_given_attrib_val(examples, 'Outlook', 'Sunny')
avt = bin_id3.construct_attrib_values_from_examples(examples, colnames[1:])
g = bin_id3.gain(outlook_sunny_examples, 'PlayTennis', 'Temperature', avt)
gt = 0.5709505944546686
err = 0.0001
print('gain(Temperature)={}'.format(g))
assert abs(g - gt) <= err
print('\n============ ID3 UT {} passed ================'.format(tn))
def test_id3_ut07(self, tn=7):
"""
Computes information gain of Wind in the examples and target_attribute PlayTennis.
"""
print('\n============ ID3 UT {} ======================='.format(tn))
examples, colnames = bin_id3.parse_csv_file_into_examples('play_tennis.csv')
avt = bin_id3.construct_attrib_values_from_examples(examples, colnames[1:])
# print('bin_id3: avt={}'.format(avt))
g = bin_id3.gain(examples, 'PlayTennis', 'Wind', avt)
gt = 0.04812703040826927
err = 0.0001
print('gain(Wind)={}'.format(g))
assert abs(g - gt) <= err
print('\n============ ID3 UT {} passed ================'.format(tn))
def test_id3_ut17(self, tn=17):
"""
Find all examples with Outlook=Overcast and compute info gain of Wind.
"""
print('\n============ ID3 UT {} ======================='.format(tn))
examples, colnames = bin_id3.parse_csv_file_into_examples('play_tennis.csv')
outlook_over_examples = bin_id3.find_examples_given_attrib_val(examples, 'Outlook', 'Overcast')
avt = bin_id3.construct_attrib_values_from_examples(examples, colnames[1:])
g = bin_id3.gain(outlook_over_examples, 'PlayTennis', 'Wind', avt)
gt = 0.0
err = 0.0001
print('gain(Wind)={}'.format(g))
assert abs(g - gt) <= err
print('\n============ ID3 UT {} passed ================'.format(tn))
def test_id3_ut03(self, tn=3):
"""
Tests bin_id3.proportion(). What proporition of examples has PlayTennis=Yes.
"""
print('\n============ ID3 UT {} ======================='.format(tn))
examples, colnames = bin_id3.parse_csv_file_into_examples('play_tennis.csv')
attribs = set(colnames[1:])
attrib = 'PlayTennis'
attrib_val = 'Yes'
p = bin_id3.proportion(examples, attrib, attrib_val)
print('proportion({}, {}) = {}'.format(attrib, attrib_val, p))
gt = 9.0/14
err = 0.0001
assert abs(p - gt) <= err
print('\n============ ID3 UT {} passed ================'.format(tn))
def test_id3_ut05(self, tn=5):
"""
- Finds the examples with Outlook=Sunny;
- In the examples with Outlook=Sunny finds all examples with PlayTennis=Yes;
- In the examples with Outlook=Sunny finds all examples with PlayTennis=No.
"""
print('\n============ ID3 UT {} ======================='.format(tn))
examples, colnames = bin_id3.parse_csv_file_into_examples('play_tennis.csv')
outlook_sunny_examples = bin_id3.find_examples_given_attrib_val(examples, 'Outlook', 'Sunny')
print('Examples with Outlook=Sunny:\n')
for i, ex in enumerate(outlook_sunny_examples):
print('{}) {}'.format(i+1, ex))
assert len(bin_id3.find_examples_given_attrib_val(outlook_sunny_examples, 'PlayTennis', 'Yes')) \
== 2
assert len(bin_id3.find_examples_given_attrib_val(outlook_sunny_examples, 'PlayTennis', 'No')) \
== 3
def test_id3_ut00(self, tn=0):
"""
Parses csv file play_tennis.csv into examples and column names.
Displays all examples.
"""
print('\n============ ID3 UT {} ======================='.format(tn))
examples, colnames = bin_id3.parse_csv_file_into_examples('play_tennis.csv')
attribs = set(colnames[1:])
print('attribs = {}'.format(attribs))
assert len(examples) == 14
print('Examples:\n')
for i, ex in enumerate(examples):
print('{}) {}'.format(i+1, ex))
print('\nColumn names:')
for i, cn in enumerate(colnames):
print('{}) {}'.format(i+1, cn))
print('\n============ ID3 UT {} passed ================'.format(tn))
def test_id3_ut21(self, tn=21):
"""
Find best (i.e., highest info gain) attribute in original Examples and display its
info gain and the gains all the other attributes.
"""
print('\n============ ID3 UT {} ======================='.format(tn))
examples, colnames = bin_id3.parse_csv_file_into_examples('play_tennis.csv')
attribs = set(colnames[1:])
avt = bin_id3.construct_attrib_values_from_examples(examples, colnames[1:])
ba, bag, gs = bin_id3.find_best_attribute(examples, 'PlayTennis', attribs, avt)
assert ba == 'Outlook'
gt = 0.2467498197744391
err = 0.0001
assert abs(bag - gt) <= err
bin_id3.display_info_gains(gs)
print('Best Attribute = {}'.format(ba))
print('Best Gain = {}'.format(bag))
print('\n============ ID3 UT {} passed ================'.format(tn))
def test_id3_ut06b(self, tn=6):
"""
Let S =
1) {'Day': 'D1', 'Outlook': 'Sunny', 'Temperature': 'Hot', 'Humidity': 'High', 'Wind': 'Weak', 'PlayTennis': 'No'}
2) {'Day': 'D2', 'Outlook': 'Sunny', 'Temperature': 'Hot', 'Humidity': 'High', 'Wind': 'Strong', 'PlayTennis': 'No'}
3) {'Day': 'D8', 'Outlook': 'Sunny', 'Temperature': 'Mild', 'Humidity': 'High', 'Wind': 'Weak', 'PlayTennis': 'No'}
4) {'Day': 'D9', 'Outlook': 'Sunny', 'Temperature': 'Cool', 'Humidity': 'Normal', 'Wind': 'Weak', 'PlayTennis': 'Yes'}
5) {'Day': 'D11', 'Outlook': 'Sunny', 'Temperature': 'Mild', 'Humidity': 'Normal', 'Wind': 'Strong', 'PlayTennis': 'Yes'}
Compute Entropy(S) w/ respect to PlayTennis;
"""
print('\n============ ID3 UT {}b ======================='.format(tn))
examples, colnames = bin_id3.parse_csv_file_into_examples('play_tennis.csv')
outlook_sunny_examples = bin_id3.find_examples_given_attrib_val(examples, 'Outlook', 'Sunny')
avt = bin_id3.construct_attrib_values_from_examples(examples, colnames[1:])
H = bin_id3.entropy(outlook_sunny_examples, 'PlayTennis', avt)
gH = 0.9709505944546686
err = 0.0001
print('Entropy(S)={}'.format(H))
assert abs(H - gH) <= err
print('\n============ ID3 UT {}b pass =================='.format(tn))
def test_id3_ut19(self, tn=19):
"""
Let S =
1) {'Day': 'D1', 'Outlook': 'Sunny', 'Temperature': 'Hot', 'Humidity': 'High', 'Wind': 'Weak', 'PlayTennis': 'No'}
2) {'Day': 'D2', 'Outlook': 'Sunny', 'Temperature': 'Hot', 'Humidity': 'High', 'Wind': 'Strong', 'PlayTennis': 'No'}
3) {'Day': 'D8', 'Outlook': 'Sunny', 'Temperature': 'Mild', 'Humidity': 'High', 'Wind': 'Weak', 'PlayTennis': 'No'}
4) {'Day': 'D9', 'Outlook': 'Sunny', 'Temperature': 'Cool', 'Humidity': 'Normal', 'Wind': 'Weak', 'PlayTennis': 'Yes'}
5) {'Day': 'D11', 'Outlook': 'Sunny', 'Temperature': 'Mild', 'Humidity': 'Normal', 'Wind': 'Strong', 'PlayTennis': 'Yes'}
Compute
1) Entropy(S);
2) Gain(Temperature);
3) Gain(Humidity);
4) Gain(Wind);
"""
print('\n============ ID3 UT {} ======================='.format(tn))
examples, colnames = bin_id3.parse_csv_file_into_examples('play_tennis.csv')
outlook_sunny_examples = bin_id3.find_examples_given_attrib_val(examples, 'Outlook', 'Sunny')
avt = bin_id3.construct_attrib_values_from_examples(examples, colnames[1:])
H = bin_id3.entropy(outlook_sunny_examples, 'PlayTennis', avt)
gH = 0.9709505944546686
err = 0.0001
print('Entropy(S)={}'.format(H))
assert abs(H - gH) <= err
g = bin_id3.gain(outlook_sunny_examples, 'PlayTennis', 'Temperature', avt)
gt = 0.5709505944546686
err = 0.0001
print('gain(Temperature)={}'.format(g))
assert abs(g-gt) <= err
g = bin_id3.gain(outlook_sunny_examples, 'PlayTennis', 'Humidity', avt)
gt = 0.9709505944546686
err = 0.0001
print('gain(Humidity)={}'.format(g))
assert abs(g-gt) <= err
g = bin_id3.gain(outlook_sunny_examples, 'PlayTennis', 'Wind', avt)
gt = 0.01997309402197489
err = 0.0001
print('gain(Wind)={}'.format(g))
assert abs(g-gt) <= err
print('\n============ ID3 UT {} passed ================'.format(tn))
