def test_matrix_confusion(self):
sc = Score(tags)
sc.fit(y, y)
cm = sc.matrix_confusion()
print(cm)
sc = Score(tags)
sc.fit(y, y_bad)
cm2 = sc.matrix_confusion()
print(cm2)
eq = np.min(np.min(cm == cm2))
self.assertEqual(eq, 0)
same_amount = np.min(cm.sum() == cm2.sum())
self.assertEqual(same_amount, 1)
def confusion(self, y_hat, y):
"""
:param y_hat:
:type y_hat:
:param y:
:type y:
:return:
:rtype:
"""
assert isinstance(y_hat, pd.Series)
assert isinstance(y, pd.Series)
a = y.values.reshape(-1)
b = pd.Series(a)
# c = b.drop(['<PAD>', '*', '<STOP>', ','])
# print(c)
roll_y = pd.Series(
y.values.reshape(-1)) #.drop(['<PAD>', '*', '<STOP>', ','])
roll_y_hat = pd.Series(
y_hat.values.reshape(-1)) #.drop(['<PAD>', '*', '<STOP>', ','])
most_reacuent_tags = self.tag_corpus[2:12]
sc = Score(most_reacuent_tags)
sc.fit(roll_y, roll_y_hat)
return sc.matrix_confusion()
def test_all_data(self):
data = PreprocessTags().load_data(r'..\data\test.wtag')
y_hat = data.y
y = data.y
roll_y = pd.Series(y.values.reshape(-1))
roll_y_hat = pd.Series(y_hat.values.reshape(-1))
index = pd.value_counts(y.values.reshape(-1)).index
most_reacuent_tags = pd.Series(index,
index=index).drop(['<STOP>', '*'])[:10]
sc = Score(most_reacuent_tags)
sc.fit(roll_y, roll_y_hat)
cm = sc.matrix_confusion()
acc_dict = sc.acc_per_tag(y, y_hat)
print(acc_dict)
print(cm)
