def test_show_fields(newsgroups_train):
docs, y, target_names = newsgroups_train
vec = TfidfVectorizer()
clf = LogisticRegression(C=0.1)
X = vec.fit_transform(docs)
clf.fit(X, y)
expl = explain_weights_sklearn(clf, vec)
text, html = format_as_all(expl, clf)
assert 'Caveats' in text
assert 'Caveats' in html
assert '<BIAS>' in text
assert 'BIAS' in html
text, html = format_as_all(expl, clf, show=fields.WEIGHTS)
assert 'Caveats' not in text
assert 'Caveats' not in html
assert '<BIAS>' in text
assert 'BIAS' in html
text, html = format_as_all(expl, clf, show=fields.INFO)
assert 'Caveats' in text
assert 'Caveats' in html
assert '<BIAS>' not in text
assert 'BIAS' not in html
def test_repr_html(boston_train):
X, y, feature_names = boston_train
reg = LinearRegression()
reg.fit(X, y)
res = explain_weights_sklearn(reg)
html = res._repr_html_()
assert 'LinearRegression' not in html
assert 'BIAS' in html
def test_show_feature_values():
clf = LinearRegression()
clf.fit(np.array([[1, 0.1], [0, 0]]), np.array([0, 1]))
res = explain_weights_sklearn(clf)
for expl in format_as_all(res, clf, show_feature_values=True):
assert 'Value' not in expl
assert 'Weight' in expl
assert 'x0' in expl
res = explain_prediction_sklearn(clf, np.array([1.52, 0.5]))
for expl in format_as_all(res, clf, show_feature_values=True):
assert 'Contribution' in expl
assert 'Value' in expl
assert 'x0' in expl
assert '1.52' in expl
for expl in format_as_all(res, clf, show_feature_values=False):
assert 'Contribution' in expl
assert 'Value' not in expl
assert 'x0' in expl
assert '1.52' not in expl
def test_highlight_spaces(boston_train, hl_spaces, add_invisible_spaces):
reg = LinearRegression()
X, y, feature_names = boston_train
# last is left unmodified to check that we are doing "any", not "all"
modified_feature_names = \
[('{} ' if add_invisible_spaces else 'A {}').format(name)
for name in feature_names[:-1]] + [feature_names[-1]]
reg.fit(X, y)
res = explain_weights_sklearn(reg,
feature_names=modified_feature_names,
top=len(feature_names) + 1)
expls = format_as_all(res, reg, highlight_spaces=hl_spaces)
for is_html, expl in enumerate(expls):
print(expl)
if hl_spaces == False or (hl_spaces is None
and not add_invisible_spaces):
for f in modified_feature_names:
assert f in expl
else:
if not add_invisible_spaces:
assert hl_spaces
for f in modified_feature_names[:-1]:
assert f not in expl
for f in feature_names[:-1]:
if is_html:
if add_invisible_spaces:
assert re.search(f + '<span.*title="A space symbol"',
expl)
else:
assert re.search(
'A<span.*title="A space symbol".*>' + f, expl)
else:
if add_invisible_spaces:
assert (f + _SPACE) in expl
else:
assert ('A' + _SPACE + f) in expl
def test_repr_html_error():
reg = BaseEstimator()
res = explain_weights_sklearn(reg)
html = res._repr_html_()
assert 'BaseEstimator' in html