def test_prediction_quality_against_benchmark():
# Given
train_data = load_dataset('train.csv')
input_df = train_data.drop(config.TARGET, axis=1)
output_df = train_data[config.TARGET]
benchmark_flexibility = 50000
benchmark_lower_boundary = (
round(output_df.iloc[0], ndigits=-4) - benchmark_flexibility
) # 210000 - 50000 = 160000
benchmark_upper_boundary = (
round(output_df.iloc[0], ndigits=-4) + benchmark_flexibility
) # 210000 + 50000 = 260000
multiple_test_json = input_df.to_json(orient='records')
# When
subject = make_prediction(input_data=multiple_test_json)
# Then
assert subject is not None
assert isinstance(subject.get('predictions')[0], float)
value = math.ceil(subject.get('predictions')[0])
assert value > benchmark_lower_boundary
assert value < benchmark_upper_boundary
def test_pipeline_drops_unnecessary_features():
# Given
test_data = load_dataset('train.csv')
X_train, X_test, y_train, y_test = train_test_split(
test_data, test_data[config.TARGET], test_size=0.1, random_state=0)
assert len(config.FEATURES) != len(X_train.columns)
X_transformed, _ = price_pipe._fit(X_train, y_train)
assert len(X_transformed[0]) == len(config.FEATURES)
def test_make_multiple_predictions():
# Given
test_data = load_dataset('test.csv')
original_data_length = len(test_data)
multiple_test_json = test_data.to_json(orient='records')
# When
subject = make_prediction(input_data=multiple_test_json)
# Then
assert subject is not None
assert len(subject.get('predictions')) != original_data_length
def test_transformer_drops_unnecessary_features():
test_data = load_dataset('train.csv')
X_train, X_test, y_train, y_test = train_test_split(
test_data, test_data[config.TARGET], test_size=0.1, random_state=0)
transformer = pp.KeepColumnsTransformer(variables=config.FEATURES, )
assert len(config.FEATURES) != len(X_train.columns)
X_transformed = transformer.transform(X_train)
assert len(X_transformed.columns) == len(config.FEATURES)
def test_pipeline_transform_min_max_features():
# Given
test_data = load_dataset('train.csv')
X_train, X_test, y_train, y_test = train_test_split(
test_data, test_data[config.TARGET], test_size=0.1, random_state=0)
X_transformed, _ = price_pipe._fit(X_train, y_train)
for x in X_transformed:
for v in x:
assert 0.0 <= v <= 1.0
def test_prediction_endpoint_returns_prediction(flask_test_client):
test_data = load_dataset(file_name=model_config.TESTING_DATA_FILE)
post_json = test_data[0:5].to_json(orient='records')
# When
response = flask_test_client.post('/v1/predict',
json=post_json)
# Then
assert response.status_code == 200
response_json = json.loads(response.data)
prediction = response_json['predictions']
assert len(prediction) == 5
def test_make_single_prediction():
# Given
test_data = load_dataset('test.csv')
single_test_json = test_data[0:1].to_json(orient='records')
# When
subject = make_prediction(input_data=single_test_json)
# Then
assert subject is not None
assert isinstance(subject.get('predictions')[0], float)
print(math.ceil(subject.get('predictions')[0]))
assert math.ceil(subject.get('predictions')[0]) == 112964
def run_training():
print("training model")
data = load_dataset(file_name=config.TRAINING_DATA_FILE)
X_train, X_test, y_train, y_test = train_test_split(data[config.FEATURES],
data[config.TARGET],
test_size=0.1,
random_state=0)
y_train = np.log(y_train)
pipeline.price_pipe.fit(X_train[config.FEATURES], y_train)
save_pipeline(pipeline_to_persist=pipeline.price_pipe)
print("training finished")