def test_determine_survival():
df = load_dataset(os.path.join(data_dir, 'train.csv'),
os.path.join(data_dir, 'test.csv'))
prob = determine_survival(df, 20, age=33, sex="male")
assert (pytest.approx(prob, 0.01) == 0.1914)
prob = determine_survival(df, 20, age=150, sex="male")
assert (prob == 0)
def test_create_new_features():
df = load_dataset(os.path.join(data_dir, 'train.csv'),
os.path.join(data_dir, 'test.csv'))
df_new_features = create_new_features(df)
assert (pytest.approx(np.exp(df_new_features.Fare_scaled).mean(),
0.01) == 34.5500)
assert (pytest.approx(df_new_features.Age_log.sum(), 0.01) == 3355.0958)
assert (df_new_features.Sex.sum() == 466)
def test_loading():
df = load_dataset(os.path.join(data_dir, 'train.csv'),
os.path.join(data_dir, 'test.csv'))
assert (isinstance(df, pd.DataFrame))
assert (df.shape == (1309, 10))
assert (df.index.tolist() == list(range(1309)))
assert (set(df['Label'].unique().tolist()) == {'Train', 'Test'})
assert (df.Label.value_counts().Train == 891)
def test_normalise():
df = load_dataset(os.path.join(data_dir, 'train.csv'),
os.path.join(data_dir, 'test.csv'))
scaled_df = normalise(df, "Fare") # 0.1708061762649229 pro col="Fare"
scaled_mean = scaled_df.loc[scaled_df.Pclass.eq(1), ["Fare"]].mean()[0]
assert (pytest.approx(scaled_mean, 0.01) == 0.1708)
scaled_mean = scaled_df.loc[scaled_df.Pclass.eq(2), ["Fare"]].mean()[0]
assert (pytest.approx(scaled_mean, 0.01) == 0.2881)
scaled_mean = scaled_df.loc[scaled_df.Pclass.eq(3), ["Fare"]].mean()[0]
assert (pytest.approx(scaled_mean, 0.01) == 0.1912)
def test_miss_data():
df = load_dataset(os.path.join(data_dir, 'train.csv'),
os.path.join(data_dir, 'test.csv'))
df_miss = get_missing_values(df)
miss_mean = df_miss.Total.sum()
assert (miss_mean == 682)
assert (df_miss.iloc[0].Total == 418)
assert (pytest.approx(df_miss.iloc[0].Percent, 0.01) == 31.9328)
assert (df_miss.iloc[-1].Total == 0)
assert (pytest.approx(df_miss.iloc[-1].Percent, 0.01) == 0)
def test_get_survived_per_class():
df = load_dataset(os.path.join(data_dir, 'train.csv'),
os.path.join(data_dir, 'test.csv'))
survived = get_survived_per_class(df, group_by_column_name="Pclass")
assert (
survived.to_json() ==
'{"Pclass":{"0":1,"1":2,"2":3},"Survived":{"0":0.63,"1":0.47,"2":0.24}}'
)
survived = get_survived_per_class(df, group_by_column_name="Sex")
assert (survived.to_json() ==
'{"Sex":{"0":"female","1":"male"},"Survived":{"0":0.74,"1":0.19}}')
def test_outliers():
df = load_dataset(os.path.join(data_dir, 'train.csv'),
os.path.join(data_dir, 'test.csv'))
n_outliers, name_max_fare = get_outliers(df) # 171, "Ward, Miss. Anna"
assert (n_outliers == 171)
assert (name_max_fare == 'Ward, Miss. Anna')
def test_correlation():
df = load_dataset(os.path.join(data_dir, 'train.csv'),
os.path.join(data_dir, 'test.csv'))
result = get_correlation(df)
assert (pytest.approx(result, 0.01) == 0.178)
def test_substitute_value():
df = load_dataset(os.path.join(data_dir, 'train.csv'),
os.path.join(data_dir, 'test.csv'))
result = substitute_missing_values(df)
assert (pytest.approx(result.iloc[17].Age, 0.01) == 29.8811)
assert (pytest.approx(result.iloc[1043].Fare, 0.01) == 14.4542)
