def test_indexname(self):
ale_eff = aleplot_1D_continuous(X=self.X,
model=self.model,
feature="x1",
grid_size=5,
include_CI=False)
self.assertEqual(ale_eff.index.name, "x1")
def test_1D_continuous_rug_plot(self):
ale_eff = aleplot_1D_continuous(
X=self.X_cleaned,
model=self.model,
feature="x1",
grid_size=5,
include_CI=True,
)
fig, ax = plot_1D_continuous_eff(ale_eff, self.X_cleaned)
## the rug
rug_plot_data = ax.lines[1].get_xydata()
# a line for each data point in X
self.assertEqual(rug_plot_data.shape[0], self.X_cleaned.shape[0])
# y position is always at the lowest eff value (including the values
# of the confidence interval)
self.assertCountEqual(
np.unique(rug_plot_data[:, 1]),
[ale_eff.drop("size", axis=1, inplace=False).min().min()],
)
# x position should always be plotted within the bin it belongs to
# (less than the upper bin limit and more than the lower bin limit)
self.assertTrue(
np.all(ale_eff.index[pd.cut(
self.X_cleaned["x1"], ale_eff.index,
include_lowest=True).cat.codes + 1] > rug_plot_data[:, 0])
and np.all(ale_eff.index[pd.cut(
self.X_cleaned["x1"], ale_eff.index,
include_lowest=True).cat.codes] < rug_plot_data[:, 0]))
def test_binsizes(self):
ale_eff = aleplot_1D_continuous(X=self.X,
model=self.model,
feature="x1",
grid_size=5,
include_CI=False)
self.assertCountEqual(ale_eff.loc[:, "size"],
[0.0, 40.0, 40.0, 40.0, 40.0, 40.0])
def test_outputshape_noCI(self):
ale_eff = aleplot_1D_continuous(X=self.X,
model=self.model,
feature="x1",
grid_size=5,
include_CI=False)
self.assertEqual(ale_eff.shape, (6, 2))
self.assertCountEqual(ale_eff.columns, ["eff", "size"])
def test_outputshape_withCI(self):
ale_eff = aleplot_1D_continuous(
X=self.X,
model=self.model,
feature="x1",
grid_size=5,
include_CI=True,
C=0.9,
)
self.assertEqual(ale_eff.shape, (6, 4))
self.assertCountEqual(ale_eff.columns,
["eff", "size", "lowerCI_90%", "upperCI_90%"])
def test_effvalues(self):
ale_eff = aleplot_1D_continuous(X=self.X,
model=self.model,
feature="x1",
grid_size=5,
include_CI=False)
self.assertCountEqual(
np.round(ale_eff.loc[:, "eff"], 8),
[
-0.35570033, -0.16996644, -0.19291121, 0.10414799, 0.24730329,
0.37855307
],
)
def test_1D_continuous_line_plot(self):
ale_eff = aleplot_1D_continuous(X=self.X,
model=self.model,
feature="x1",
grid_size=5,
include_CI=True)
fig, ax = plot_1D_continuous_eff(ale_eff, self.X)
## effect line
eff_plt_data = ax.lines[0].get_xydata()
# the x values should be the bins
self.assertCountEqual(eff_plt_data[:, 0], ale_eff.index)
# the y values should be the effect
self.assertCountEqual(eff_plt_data[:, 1], ale_eff.eff)
def test_effvalues(self):
ale_eff = aleplot_1D_continuous(
X=self.X_cleaned,
model=self.model,
feature="x1",
grid_size=5,
include_CI=False,
)
self.assertCountEqual(
np.round(ale_eff.loc[:, "eff"], 8),
[
-0.3302859, -0.25946135, -0.03809224, 0.03292833, 0.27153761,
0.3164612
],
)
def test_1D_continuous_ci_plot(self):
ale_eff = aleplot_1D_continuous(X=self.X,
model=self.model,
feature="x1",
grid_size=5,
include_CI=True)
fig, ax = plot_1D_continuous_eff(ale_eff, self.X)
ci_plot_data = pd.DataFrame(
ax.collections[0].get_paths()
[0].vertices).drop_duplicates().groupby(0).agg(['min', 'max'])
ci_plot_data.index.name = 'x1'
ci_plot_data.columns = ['lowerCI_95%', 'upperCI_95%']
self.assertTrue(
np.all(
ale_eff.loc[ale_eff.index[1]:,
['lowerCI_95%', 'upperCI_95%']] == ci_plot_data))
def test_bins(self):
ale_eff = aleplot_1D_continuous(X=self.X,
model=self.model,
feature="x1",
grid_size=5,
include_CI=False)
self.assertCountEqual(
ale_eff.index,
[
0.0013107121819164735,
0.21205399821897986,
0.3905585553320686,
0.5561380185409515,
0.7797798975036754,
0.9986526271693825,
],
)
def test_1D_continuous_ci_plot(self):
ale_eff = aleplot_1D_continuous(
X=self.X_cleaned,
model=self.model,
feature="x1",
grid_size=5,
include_CI=True,
)
fig, ax = plot_1D_continuous_eff(ale_eff, self.X_cleaned)
ci_plot_data = (pd.DataFrame(
ax.collections[0].get_paths()
[0].vertices).drop_duplicates().groupby(0).agg(["min", "max"]))
ci_plot_data.index.name = "x1"
ci_plot_data.columns = ["lowerCI_95%", "upperCI_95%"]
self.assertTrue(
np.all(
ale_eff.loc[ale_eff.index[1]:,
["lowerCI_95%", "upperCI_95%"]] == ci_plot_data))
def test_CIvalues(self):
ale_eff = aleplot_1D_continuous(
X=self.X,
model=self.model,
feature="x1",
grid_size=5,
include_CI=True,
C=0.9,
)
# assert that the first bin do not have a CI
self.assertTrue(np.isnan(ale_eff.loc[ale_eff.index[0], "lowerCI_90%"]))
self.assertTrue(np.isnan(ale_eff.loc[ale_eff.index[0], "upperCI_90%"]))
# check the values of the CI
self.assertCountEqual(
np.round(ale_eff.loc[ale_eff.index[1]:, "lowerCI_90%"], 8),
[-0.21966029, -0.27471201, -0.01534647, 0.20038572, 0.30378132],
)
self.assertCountEqual(
np.round(ale_eff.loc[ale_eff.index[1]:, "upperCI_90%"], 8),
[-0.12027259, -0.11111041, 0.22364245, 0.29422086, 0.45332483],
)
def test_CIvalues(self):
ale_eff = aleplot_1D_continuous(
X=self.X_cleaned,
model=self.model,
feature="x1",
grid_size=5,
include_CI=True,
C=0.9,
)
# assert that the first bin do not have a CI
self.assertTrue(np.isnan(ale_eff.loc[ale_eff.index[0], "lowerCI_90%"]))
self.assertTrue(np.isnan(ale_eff.loc[ale_eff.index[0], "upperCI_90%"]))
# check the values of the CI
self.assertCountEqual(
np.round(ale_eff.loc[ale_eff.index[1]:, "lowerCI_90%"], 8),
[-0.37210104, -0.08077478, -0.00175768, 0.20772107, 0.24621853],
)
self.assertCountEqual(
np.round(ale_eff.loc[ale_eff.index[1]:, "upperCI_90%"], 8),
[-0.14682166, 0.00459031, 0.06761434, 0.33535415, 0.38670386],
)
def test_1D_continuous_rug_plot(self):
ale_eff = aleplot_1D_continuous(X=self.X,
model=self.model,
feature="x1",
grid_size=5,
include_CI=True)
fig, ax = plot_1D_continuous_eff(ale_eff, self.X)
## the rug
rug_plot_data = ax.lines[1].get_xydata()
# a line for each data point in X
self.assertEqual(rug_plot_data.shape[0], self.X.shape[0])
# y position is always at the lower eff value
self.assertCountEqual(np.unique(rug_plot_data[:, 1]),
[ale_eff.eff.min()])
# x position should always be plotted within the bin it belongs to
# (less than the upper bin limit and more than the lower bin limit)
self.assertTrue(
np.all(ale_eff.index[pd.cut(
self.X["x1"], ale_eff.index, include_lowest=True).cat.codes +
1] > rug_plot_data[:, 0])
and np.all(ale_eff.index[pd.cut(
self.X["x1"], ale_eff.index, include_lowest=True).cat.codes] <
rug_plot_data[:, 0]))
def test_exceptions(self):
# dataset should be compatible with the model
with self.assertRaises(Exception) as mod_ex_2:
aleplot_1D_continuous(self.X, self.model, "x1")
mod_ex_msg = "Please check that your model is fitted, and accepts X as input."
self.assertEqual(mod_ex_2.exception.args[0], mod_ex_msg)