def _compute_lace_step(self,
encoded_instance,
k, target_class_frequency,
target_class_index, class_prob, single_attribute_differences):
# Generate the neighborhood of the instance, classify it, and return the rules created by L3
l3clf = L3Classifier(min_sup=self.min_sup)
rules = _create_locality_and_get_rules(self.train_dataset, self.nbrs, encoded_instance, k,
self.clf, l3clf)
# For each rule, calculate the prediction difference for the its attributes
difference_map = {}
for rule in rules:
rule_key = ",".join(map(str, rule))
difference_map[rule_key] = _compute_prediction_difference_subset(
self.train_dataset, encoded_instance, rule,
self.clf, target_class_index)
# Compute the approximation error
_, error, _ = _compute_approximation_error(
target_class_frequency,
class_prob,
single_attribute_differences,
rules,
difference_map)
return difference_map, error
def test_save_load(dataset_X_y):
X, y = dataset_X_y
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.33,
random_state=42)
clf = L3Classifier().fit(X_train, y_train)
# test dump with joblib pre-predict
dump(clf, "clf_pre_predict.joblib")
y_pred = clf.predict(X_test)
assert y_pred.shape[0] == X_test.shape[0]
assert len(clf.labeled_transactions_) == X_test.shape[0]
# test dump with joblib post-predict
dump(clf, "clf.joblib")
clf_l = load("clf.joblib")
assert len(clf.lvl1_rules_) == len(clf_l.lvl1_rules_)
# test dump with pickle
with open("clf.pickle", "wb") as fp:
pickle.dump(clf, fp)
with open("clf.pickle", "rb") as fp:
clf_l = pickle.load(fp)
assert len(clf.lvl2_rules_) == len(clf_l.lvl2_rules_)
def test_grid_search(dataset_X_y, get_param_grid):
X, y = dataset_X_y
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.33,
random_state=42)
clf = GridSearchCV(L3Classifier(), get_param_grid, n_jobs=-1)
clf.fit(X, y)
print(clf.best_estimator_)
print(clf.best_score_)
def test_training_files(dataset_X_y):
X, y = dataset_X_y
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.33,
random_state=42)
clf = L3Classifier().fit(X_train, y_train, remove_files=False)
train_dir = clf.current_token_
files = [
f for f in os.listdir(train_dir)
if f.startswith(f"{clf.current_token_}")
]
assert len(files) == 7 # all the stuff left by L3
def test_save_human_readable(dataset_X_y):
X, y = dataset_X_y
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.33,
random_state=42)
clf = L3Classifier().fit(X_train, y_train, save_human_readable=True)
train_dir = clf.current_token_
files = [
f for f in os.listdir(train_dir)
if f.startswith(f"{clf.current_token_}")
]
assert len(files) == 2 # level 1 and level 2
def test_fit_predict(dataset_X_y):
X, y = dataset_X_y
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.33,
random_state=42)
clf = L3Classifier().fit(X_train, y_train)
y_pred = clf.predict(X_test)
assert y_pred.shape[0] == X_test.shape[0]
assert len(clf.labeled_transactions_) == X_test.shape[0]
print(clf.labeled_transactions_[1].matched_rules,
clf.labeled_transactions_[1].used_level)
print(len([t for t in clf.labeled_transactions_ if t.used_level == -1]))
def test_leve1_modifier(dataset_X_y):
X, y = dataset_X_y
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.33,
random_state=42)
clf = L3Classifier(rule_sets_modifier='level1').fit(X, y)
assert clf.n_lvl2_rules_ == 0 and len(clf.lvl2_rules_) == 0
y_pred = clf.predict(X_test)
assert y_pred.shape[0] == X_test.shape[0]
assert len(clf.labeled_transactions_) == X_test.shape[0]
print(clf.labeled_transactions_[1].matched_rules,
clf.labeled_transactions_[1].used_level)
assert len([t for t in clf.labeled_transactions_
if t.used_level == 2]) == 0