def cosine_test(train_text, test_text, num1, num2):
if not train_text or not test_text:
return 0.0
tfidf_vectorizer = TfidfVectorizer()
tfidf_matrix = tfidf_vectorizer.fit_transform([train_text.lower(), test_text.lower()])
result = (tfidf_matrix * tfidf_matrix.T).A[0, 1]
return balance_result(num1, num2, False, result)
def abbr_test(train_examples, test_examples, num1, num2):
# if testExamples is a string, perform metadata abbr test (label thing). Else do the normal one
train_example_set = set(train_examples)
count_matches = 0
if isinstance(test_examples, str):
patterns = get_abbr_patterns(test_examples)
for pattern in patterns:
if pattern.match(train_examples):
count_matches += 1
break
return count_matches
else:
test_example_set = set(test_examples)
if len(test_example_set) > 50 or len(train_example_set) > 50:
return 0.0
for test_example in test_example_set:
if not test_example.isupper():
continue
patterns = get_abbr_patterns(test_example)
found = False
for pattern in patterns:
for train_example in train_example_set:
if pattern.match(train_example):
count_matches += 1
found = True
break
if found:
break
result = count_matches * 2.0 / (len(train_example_set) + len(test_example_set))
return balance_result(num1, num2, False, result)
def abbr_test(train_examples, test_examples, num1, num2):
# if testExamples is a string, perform metadata abbr test (label thing). Else do the normal one
train_example_set = set(train_examples)
count_matches = 0
if isinstance(test_examples, str):
patterns = get_abbr_patterns(test_examples)
for pattern in patterns:
if pattern.match(train_examples):
count_matches += 1
break
return count_matches
else:
test_example_set = set(test_examples)
if len(test_example_set) > 50 or len(train_example_set) > 50:
return 0.0
for test_example in test_example_set:
if not test_example.isupper():
continue
patterns = get_abbr_patterns(test_example)
found = False
for pattern in patterns:
for train_example in train_example_set:
if pattern.match(train_example):
count_matches += 1
found = True
break
if found:
break
result = count_matches * 2.0 / (len(train_example_set) +
len(test_example_set))
return balance_result(num1, num2, False, result)
def cosine_test(train_text, test_text, num1, num2):
if not train_text or not test_text:
return 0.0
tfidf_vectorizer = TfidfVectorizer()
tfidf_matrix = tfidf_vectorizer.fit_transform(
[train_text.lower(), test_text.lower()])
result = (tfidf_matrix * tfidf_matrix.T).A[0, 1]
return balance_result(num1, num2, False, result)
def mann_whitney_u_test(train_examples, test_examples, num1, num2):
try:
if len(train_examples) > 1 and len(test_examples) > 1:
result = mannwhitneyu(train_examples, test_examples)[1]
return balance_result(num1, num2, True, result)
return 0
except ValueError as e:
logging.warn("IGNORE EXCEPTION: %s", str(e))
return 0
def mann_whitney_u_test(train_examples, test_examples, num1, num2):
try:
if len(train_examples) > 1 and len(test_examples) > 1:
result = mannwhitneyu(train_examples, test_examples)[1]
return balance_result(num1, num2, True, result)
return 0
except ValueError as e:
logging.warn("IGNORE EXCEPTION: %s", str(e))
return 0
def coverage_test(train_examples, test_examples, num1, num2):
if len(train_examples) > 1 and len(test_examples) > 1:
max1 = percentile(train_examples, 75)
min1 = percentile(train_examples, 25)
max2 = percentile(test_examples, 75)
min2 = percentile(test_examples, 25)
max3 = max(max1, max2)
min3 = min(min1, min2)
if min2 > max1 or min1 > max2:
return 0
elif max3 == min3:
return 0
else:
min4 = min(max1, max2)
max4 = max(min1, min2)
result = (min4 - max4) * 1.0 / (max3 - min3)
return balance_result(num1, num2, True, result)
return 0
def coverage_test(train_examples, test_examples, num1, num2):
if len(train_examples) > 1 and len(test_examples) > 1:
max1 = percentile(train_examples, 100)
min1 = percentile(train_examples, 0)
max2 = percentile(test_examples, 100)
min2 = percentile(test_examples, 0)
max3 = max(max1, max2)
min3 = min(min1, min2)
# print "max1", max1
# print "min1", min1
# print "max2", max2
# print "min2", min2
if min2 > max1 or min1 > max2:
return 0
elif max3 == min3:
return 0
else:
min4 = min(max1, max2)
max4 = max(min1, min2)
result = (min4 - max4) * 1.0 / (max3 - min3)
return balance_result(num1, num2, True, result)
return 0
def welch_test(train_examples, test_examples, num1, num2):
if len(train_examples) > 1 and len(test_examples) > 1:
print(train_examples, test_examples)
result = ttest_ind(train_examples, test_examples, False)[1]
return balance_result(num1, num2, True, result)
return 0
def kolmogorov_smirnov_test(train_examples, test_examples, num1, num2):
if len(train_examples) > 1 and len(test_examples) > 1:
result = ks_2samp(train_examples, test_examples)[1]
return balance_result(num1, num2, True, result)
return 0
def jaccard_test(train_examples, test_examples, num1, num2):
result = jaccard_similarity(train_examples, test_examples)
return balance_result(num1, num2, False, result)
def welch_test(train_examples, test_examples, num1, num2):
if len(train_examples) > 1 and len(test_examples) > 1:
print(train_examples, test_examples)
result = ttest_ind(train_examples, test_examples, False)[1]
return balance_result(num1, num2, True, result)
return 0
def kolmogorov_smirnov_test(train_examples, test_examples, num1, num2):
if len(train_examples) > 1 and len(test_examples) > 1:
result = ks_2samp(train_examples, test_examples)[1]
return balance_result(num1, num2, True, result)
return 0
def mann_whitney_u_test(train_examples, test_examples, num1, num2):
if len(train_examples) > 1 and len(test_examples) > 1:
result = mannwhitneyu(train_examples, test_examples)[1]
return balance_result(num1, num2, True, result)
return 0
def jaccard_test(train_examples, test_examples, num1, num2):
result = jaccard_similarity(train_examples, test_examples)
return balance_result(num1, num2, False, result)