def test_class_weights(self, iris_X, iris_y):
X_train, X_test, y_train, y_test = train_test_split(iris_X,
iris_y,
test_size=0.5,
random_state=42)
forest = RangerForestClassifier()
weights = {
0: 0.7,
1: 0.2,
2: 0.1,
}
forest.fit(X_train, y_train, class_weights=weights)
forest.predict(X_test)
forest = RangerForestClassifier()
m = {0: "a", 1: "b", 2: "c"}
y_train_str = [m.get(v) for v in y_train]
weights = {
"a": 0.7,
"b": 0.2,
"c": 0.1,
}
forest.fit(X_train, y_train_str, class_weights=weights)
forest.predict(X_test)
weights = {
0: 0.7,
}
with pytest.raises(ValueError):
forest.fit(X_train, y_train, class_weights=weights)
def test_predict(self, iris_X, iris_y):
rfc = RangerForestClassifier()
rfc.fit(iris_X, iris_y)
pred = rfc.predict(iris_X)
assert len(pred) == iris_X.shape[0]
# test with single record
iris_X_record = iris_X[0:1, :]
pred = rfc.predict(iris_X_record)
assert len(pred) == 1
def test_categorical_features(self, iris_X, iris_y,
respect_categorical_features):
# add a categorical feature
categorical_col = np.atleast_2d(
np.array([random.choice([0, 1]) for _ in range(iris_X.shape[0])]))
iris_X_c = np.hstack((iris_X, categorical_col.transpose()))
categorical_features = [iris_X.shape[1]]
forest = RangerForestClassifier(
respect_categorical_features=respect_categorical_features, )
if respect_categorical_features not in [
"partition", "ignore", "order"
]:
with pytest.raises(ValueError):
forest.fit(iris_X_c,
iris_y,
categorical_features=categorical_features)
return
forest.fit(iris_X_c, iris_y, categorical_features=categorical_features)
forest.predict(iris_X_c)
def test_sample_fraction(self, iris_X, iris_y):
rfc = RangerForestClassifier(sample_fraction=[0.69])
rfc.fit(iris_X, iris_y)
assert rfc.sample_fraction_ == [0.69]
rfc = RangerForestClassifier(sample_fraction=0.69)
rfc.fit(iris_X, iris_y)
assert rfc.sample_fraction_ == [0.69]
# test with single record
iris_X_record = iris_X[0:1, :]
pred = rfc.predict(iris_X_record)
assert len(pred) == 1
pred = rfc.predict_proba(iris_X_record)
assert len(pred) == 1
pred = rfc.predict_log_proba(iris_X_record)
assert len(pred) == 1
def test_accuracy(self, iris_X, iris_y):
X_train, X_test, y_train, y_test = train_test_split(iris_X, iris_y, test_size=0.33, random_state=42)
# train and test a random forest classifier
rf = RandomForestClassifier()
rf.fit(X_train, y_train)
y_pred_rf = rf.predict(X_test)
rf_acc = accuracy_score(y_test, y_pred_rf)
# train and test a ranger classifier
ra = RangerForestClassifier()
ra.fit(X_train, y_train)
y_pred_ra = ra.predict(X_test)
ranger_acc = accuracy_score(y_test, y_pred_ra)
# the accuracy should be good
assert rf_acc > 0.9
assert ranger_acc > 0.9
def RangerForestIMPL(train_embeddings, train_labels,test_embeddings,test_labels,single_label=True):
rfc = RangerForestClassifier()
rfc.fit(train_embeddings, train_labels)
predsranger = rfc.predict(test_embeddings)
if single_label==True:
return single_label_metrics(test_labels,predsranger)