def test_importance_preprocess(data, compute_backend_df):
def pre(df, target):
y = df[target]
df = df.drop(target, axis=1)
x_num = df.select_dtypes(["number"])
x_num = x_num.fillna("-1")
x_cat = df[[c for c in df.columns
if c not in x_num.columns]].astype(str)
x_cat = x_cat.fillna("")
x_cat_encoded = x_cat.apply(LabelEncoder().fit_transform)
X = pd.concat([x_num, x_cat_encoded], axis=1)
return X, y
fig = dd.importance(compute_backend_df, "d", preprocess_func=pre)
assert isinstance(fig, matplotlib.artist.Artist)
def test_top_feature(compute_backend_df):
fig = dd.importance(compute_backend_df, "d", top_features=1)
assert isinstance(fig, matplotlib.artist.Artist)
def test_importance_num_only(data, compute_backend_df):
data = compute_backend_df.select_dtypes(["number"])
rfr = RandomForestRegressor(random_state=1)
assert isinstance(
dd.importance(data, "a", estimator=rfr, return_values=True),
np.ndarray), "Importance values not a numpy array"
def test_importance_cat_only(data, compute_backend_df):
num_columns = compute_backend_df.select_dtypes(["number"]).columns.values
data = data[[c for c in data.columns if c not in num_columns]]
assert (len(dd.importance(data, "d",
return_values=True)) == data.shape[1] -
2), "Wrong size of importance values" # f is null column
def test_importance(compute_backend_df):
importance_vals = dd.importance(compute_backend_df,
"d",
return_values=True)
assert len(importance_vals) == compute_backend_df.shape[1] - 1 - 1