def test_numerical_monotonic_xy():
optb_mip = OptimalBinning2D(solver="mip",
monotonic_trend_x="descending",
monotonic_trend_y="descending")
optb_cp = OptimalBinning2D(solver="cp",
monotonic_trend_x="descending",
monotonic_trend_y="descending")
for optb in [optb_mip, optb_cp]:
optb.fit(x, y, z)
assert optb.status == "OPTIMAL"
optb.binning_table.build()
assert optb.binning_table.iv == approx(7.59474677, rel=1e-6)
optb_mip = OptimalBinning2D(solver="mip",
monotonic_trend_x="ascending",
monotonic_trend_y="ascending")
optb_cp = OptimalBinning2D(solver="cp",
monotonic_trend_x="ascending",
monotonic_trend_y="ascending")
for optb in [optb_mip, optb_cp]:
optb.fit(x, y, z)
assert optb.status == "OPTIMAL"
optb.binning_table.build()
assert optb.binning_table.iv == approx(0, rel=1e-6)
def test_numerical_regularization():
optb_mip = OptimalBinning2D(min_bin_size=0.05, solver="mip", gamma=600)
optb_cp = OptimalBinning2D(min_bin_size=0.05, solver="cp", gamma=600)
optb_mip.fit(x, y, z)
optb_cp.fit(x, y, z)
assert len(optb_mip.splits[0]) < 6
assert len(optb_cp.splits[0]) < 6
def test_numerical_min_max_n_bins():
optb_mip = OptimalBinning2D(solver="mip", min_n_bins=2, max_n_bins=5)
optb_cp = OptimalBinning2D(solver="cp", min_n_bins=2, max_n_bins=5)
for optb in [optb_mip, optb_cp]:
optb.fit(x, y, z)
assert optb.status == "OPTIMAL"
assert 2 <= len(optb.splits[0]) <= 5
assert 2 <= len(optb.splits[1]) <= 5
def test_numerical_default_solvers():
optb_mip = OptimalBinning2D(solver="mip")
optb_cp = OptimalBinning2D(solver="cp")
for optb in (optb_mip, optb_cp):
optb.fit(x, y, z)
assert optb.status == "OPTIMAL"
optb.binning_table.build()
assert optb.binning_table.iv == approx(7.63248244, rel=1e-6)
def test_numerical_strategy():
optb = OptimalBinning2D(strategy="cart")
optb.fit(x, y, z)
assert optb.status == "OPTIMAL"
optb.binning_table.build()
assert optb.binning_table.iv == approx(6.63764258, rel=1e-6)
def test_information():
optb = OptimalBinning2D(solver="cp")
with raises(NotFittedError):
optb.information()
optb.fit(x, y, z)
with raises(ValueError):
optb.information(print_level=-1)
optb.information(print_level=0)
optb.information(print_level=1)
optb.information(print_level=2)
optb = OptimalBinning2D(solver="mip")
optb.fit(x, y, z)
optb.information(print_level=2)
def test_numerical_default_transform():
optb = OptimalBinning2D()
with raises(NotFittedError):
z_transform = optb.transform(x, y)
optb.fit(x, y, z)
with raises(ValueError):
z_transform = optb.transform(x, y, metric="new_metric")
z_transform = optb.transform(x, y, metric="woe")
assert z_transform[:5] == approx(
[5.37317977, 3.51688178, 5.37317977, 0.52114951, 5.37317977], rel=1e-6)
z_transform = optb.transform(x, y, metric="event_rate", check_input=True)
assert z_transform[:5] == approx(
[0.00775194, 0.04761905, 0.00775194, 0.5, 0.00775194], rel=1e-6)
z_transform = optb.transform(x, y, metric="indices")
assert z_transform[:5] == approx([12, 3, 12, 13, 12], rel=1e-6)
z_transform = optb.transform(x, y, metric="bins")
assert z_transform[0] == '[15.05, inf) $\\cup$ [0.32, inf)'
def test_numerical_default():
optb = OptimalBinning2D()
optb.fit(x, y, z)
assert optb.status == "OPTIMAL"
with raises(TypeError):
optb.binning_table.build(show_bin_xy=1)
optb.binning_table.build()
assert optb.binning_table.iv == approx(7.63248244, rel=1e-6)
optb.binning_table.analysis()
assert optb.binning_table.gini == approx(0.96381005, rel=1e-6)
assert optb.binning_table.js == approx(0.53356918, rel=1e-6)
assert optb.binning_table.quality_score == approx(0.0, rel=1e-6)
with raises(ValueError):
optb.binning_table.plot(metric="new_metric")
optb.binning_table.plot(metric="woe", savefig="test_binning_2d_woe.png")
optb.binning_table.plot(metric="event_rate",
savefig="test_binning_2d_event_rate.png")
def test_verbose():
optb = OptimalBinning2D(verbose=True)
optb.fit(x, y, z)
assert optb.status == "OPTIMAL"
def test_numerical_default_fit_transform():
optb = OptimalBinning2D()
z_transform = optb.fit_transform(x, y, z, metric="woe")
assert z_transform[:5] == approx(
[5.37317977, 3.51688178, 5.37317977, 0.52114951, 5.37317977], rel=1e-6)
def test_params():
with raises(TypeError):
optb = OptimalBinning2D(name_x=1)
optb.fit(x, y, z)
with raises(TypeError):
optb = OptimalBinning2D(name_y=1)
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(dtype_x="nominal")
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(dtype_y="nominal")
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(prebinning_method="new_method")
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(strategy="new_strategy")
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(solver="new_solver")
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(divergence="new_divergence")
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(max_n_prebins_x=-2)
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(max_n_prebins_y=-2)
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(min_prebin_size_x=0.6)
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(min_prebin_size_y=0.6)
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(min_n_bins=-2)
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(max_n_bins=-2.2)
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(min_n_bins=3, max_n_bins=2)
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(min_bin_size=0.6)
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(max_bin_size=-0.6)
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(min_bin_size=0.5, max_bin_size=0.3)
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(min_bin_n_nonevent=-2)
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(max_bin_n_nonevent=-2)
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(min_bin_n_nonevent=3, max_bin_n_nonevent=2)
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(min_bin_n_event=-2)
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(max_bin_n_event=-2)
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(min_bin_n_event=3, max_bin_n_event=2)
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(monotonic_trend_x="new_trend")
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(monotonic_trend_y="new_trend")
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(min_event_rate_diff_x=-1.1)
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(min_event_rate_diff_y=1.1)
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(gamma=-0.2)
optb.fit(x, y, z)
with raises(TypeError):
optb = OptimalBinning2D(special_codes_x={1, 2, 3})
optb.fit(x, y, z)
with raises(TypeError):
optb = OptimalBinning2D(special_codes_y={1, 2, 3})
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(split_digits=9)
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(n_jobs=1.2)
optb.fit(x, y, z)
with raises(ValueError):
optb = OptimalBinning2D(time_limit=-2)
optb.fit(x, y, z)
with raises(TypeError):
optb = OptimalBinning2D(verbose=1)
optb.fit(x, y, z)