def test_data_pipeline(gamelocation):
"""Test game location."""
flow = gen_data_pipeline()
output = run_pipeline(flow=flow,
data_dir=gamelocation,
output_dir=gamelocation,
seed=42)
assert output.is_successful()
assert Path(gamelocation, "models", "train.csv").is_file()
assert Path(gamelocation, "models", "tune.csv").is_file()
assert Path(gamelocation, "models", "holdout.csv").is_file()
train = pd.read_csv(Path(gamelocation, "models", "train.csv"),
sep="|",
dtype={"GAME_ID": str})
tune = pd.read_csv(Path(gamelocation, "models", "tune.csv"),
sep="|",
dtype={"GAME_ID": str})
holdout = pd.read_csv(Path(gamelocation, "models", "holdout.csv"),
sep="|",
dtype={"GAME_ID": str})
assert len(np.unique(train["GAME_ID"].values)) == 120
assert len(np.unique(tune["GAME_ID"].values)) == 40
assert len(np.unique(holdout["GAME_ID"].values)) == 40
def test_evaluate_pipeline(mock_auc, gamelocation):
"""Test evaluate pipeline."""
mock_auc.return_value = 0.5
flow = gen_evaluate_pipeline(
step=288,
xgboost=Path(gamelocation, "models", TODAY.strftime("%Y-%m-%d"),
"xgboost", "model.pkl"),
)
output = run_pipeline(flow=flow,
data_dir=gamelocation,
output_dir=gamelocation)
assert output.is_successful()
assert Path(gamelocation, "models", TODAY.strftime("%Y-%m-%d"),
"auroc.png").is_file()
assert Path(gamelocation, "models", TODAY.strftime("%Y-%m-%d"),
"auroc_lift.png").is_file()
def test_xgboost_pipeline(mock_auc, gamelocation):
"""Test fitting XGBoost model."""
mock_auc.return_value = 0.5
# Create and run the XGBoost flow
flow = gen_xgboost_pipeline()
output = run_pipeline(flow=flow,
data_dir=gamelocation,
output_dir=gamelocation,
max_evals=5,
seed=42)
assert output.is_successful()
assert Path(gamelocation, "models", TODAY.strftime("%Y-%m-%d"), "xgboost",
"model.pkl").is_file()
assert Path(gamelocation, "models", TODAY.strftime("%Y-%m-%d"), "xgboost",
"tuning.pkl").is_file()
assert Path(gamelocation, "models", TODAY.strftime("%Y-%m-%d"), "xgboost",
"hyperparameter-tuning.png").is_file()
def test_predict_pipeline(mock_pred, gamelocation):
"""Test the prediction pipeline."""
mock_pred.return_value = pd.Series([0.5])
flow = gen_predict_pipeline()
output = run_pipeline(flow=flow,
data_dir=gamelocation,
output_dir=gamelocation,
filesystem="file",
model=str(
Path(gamelocation, "models",
TODAY.strftime("%Y-%m-%d"), "lifelines",
"model.pkl")))
assert output.is_successful()
assert Path(gamelocation, "2018-19", "survival-prediction").is_dir()
assert len(
list(
Path(gamelocation, "2018-19",
"survival-prediction").glob("data_*.csv"))) == 200
def test_lifelines_pipelines(mock_auc, gamelocation):
"""Test fitting a lifelines model."""
mock_auc.return_value = 0.5
# Create and run the flow
train = pd.read_csv(Path(gamelocation, "models", "train.csv"),
sep="|",
dtype={"GAME_ID": str})
tune = pd.read_csv(Path(gamelocation, "models", "tune.csv"),
sep="|",
dtype={"GAME_ID": str})
holdout = pd.read_csv(Path(gamelocation, "models", "holdout.csv"),
sep="|",
dtype={"GAME_ID": str})
# Drop nulls because of the weirdness with random data
train = train.dropna()
train.to_csv(Path(gamelocation, "models", "train.csv"), sep="|")
tune = tune.dropna()
tune.to_csv(Path(gamelocation, "models", "tune.csv"), sep="|")
holdout = holdout.dropna()
holdout.to_csv(Path(gamelocation, "models", "holdout.csv"), sep="|")
# Create and run the lifelines flow
flow = gen_lifelines_pipeline()
output = run_pipeline(flow=flow,
data_dir=gamelocation,
output_dir=gamelocation,
max_evals=5,
seed=42)
assert output.is_successful()
assert Path(gamelocation, "models", TODAY.strftime("%Y-%m-%d"),
"lifelines", "model.pkl").is_file()
assert Path(gamelocation, "models", TODAY.strftime("%Y-%m-%d"),
"lifelines", "tuning.pkl").is_file()
assert Path(gamelocation, "models", TODAY.strftime("%Y-%m-%d"),
"lifelines", "hyperparameter-tuning.png").is_file()