def test_split2_allclass():
train_idx, test_idx, label_idx, unlabel_idx = split_multi_label(
y=mult_y, label_shape=(instance_num, label_num),
all_class=True, split_count=split_count,
test_ratio=0.3, initial_label_rate=0.05,
saving_path=None
)
assert len(train_idx) == split_count
assert len(test_idx) == split_count
assert len(label_idx) == split_count
assert len(unlabel_idx) == split_count
for i in range(split_count):
check_index_multilabel(label_idx[i])
check_index_multilabel(unlabel_idx[i])
train = set(train_idx[i])
test = set(test_idx[i])
assert len(label_idx[i]) == len(integrate_multilabel_index(label_idx[i], label_size=label_num))
# validity
lab = set([j[0] for j in label_idx[i]])
unl = set([j[0] for j in unlabel_idx[i]])
traintest = train.union(test)
labun = lab.union(unl)
assert len(test) == round(0.3 * instance_num)
assert len(lab) == round(0.05 * len(train))
assert traintest == set(range(instance_num))
assert labun == train
def _get_cond_instance(self, cond):
"""Return the indexes of instances according to the cond.
cond = 0: return the instances which are unbroken.
cond = 1: return the instances which have missing entries.
"""
tmp = integrate_multilabel_index(self.index,
label_size=self._label_size,
check_arr=False)
if cond == 0:
return [tp[0] for tp in tmp if len(tp) == 1]
else:
return [tp[0] for tp in tmp if len(tp) > 1]
def test_split3_all_features():
train_idx, test_idx, label_idx, unlabel_idx = split_features(feature_matrix=X, feature_matrix_shape=X.shape,
test_ratio=0.3, missing_rate=0.2,
split_count=split_count,
all_features=True,
saving_path=None)
assert len(train_idx) == split_count
assert len(test_idx) == split_count
assert len(label_idx) == split_count
assert len(unlabel_idx) == split_count
for i in range(split_count):
train = set(train_idx[i])
test = set(test_idx[i])
traintest = train.union(test)
# validity
assert len(flattern_multilabel_index(index_arr=unlabel_idx[i], label_size=feature_num)) == round(
0.2 * len(train) * X.shape[1])
assert len(test) == round(0.3 * instance_num)
assert traintest == set(range(instance_num))
assert len(
[j[0] for j in integrate_multilabel_index(label_idx[i] + unlabel_idx[i], label_size=feature_num)]) == len(
train_idx[i])