def test_accuracy_improvement():
base_model = XGBSEKaplanTree()
base_model.fit(X_train, y_train)
bootstrap = XGBSEBootstrapEstimator(base_model)
bootstrap.fit(X_train, y_train)
cind_base = concordance_index(y_test, base_model.predict(X_test))
cind_boots = concordance_index(y_test, bootstrap.predict(X_test))
assert cind_boots > cind_base
def test_concordance_index():
assert concordance_index(y_train, km_survival) == 0.5
assert concordance_index(y_test, preds) > 0.5
assert np.isclose(
concordance_index(y_test, T_test.values, risk_strategy="precomputed"),
0,
atol=0.02,
)
assert np.isclose(
concordance_index(y_test, -T_test.values, risk_strategy="precomputed"),
1,
atol=0.02,
)
def test_survival_curve_tree():
xgbse = XGBSEKaplanTree()
xgbse.fit(X_train, y_train)
preds = xgbse.predict(X_test)
cindex = concordance_index(y_test, preds)
assert_survival_curve(xgbse, X_test, preds, cindex)
def test_survival_curve_without_early_stopping(model):
xgbse = model()
xgbse.fit(
X_train,
y_train,
)
preds = xgbse.predict(X_test)
cindex = concordance_index(y_test, preds)
assert_survival_curve(xgbse, X_test, preds, cindex)
def test_survival_curve(model):
xgbse = model()
xgbse.fit(
X_train,
y_train,
num_boost_round=1000,
validation_data=(X_valid, y_valid),
early_stopping_rounds=10,
verbose_eval=0,
)
preds = xgbse.predict(X_test)
cindex = concordance_index(y_test, preds)
assert_survival_curve(xgbse, X_test, preds, cindex)