def test_n_estimators(make_whas500):
whas500 = make_whas500(with_mean=False,
with_std=False,
to_numeric=True)
base_estimators = [('gbm',
ComponentwiseGradientBoostingSurvivalAnalysis()),
('svm', FastSurvivalSVM())]
meta = EnsembleSelection(base_estimators,
scorer=score_cindex,
n_estimators=0)
with pytest.raises(ValueError,
match="n_estimators must not be zero or negative"):
meta.fit(whas500.x, whas500.y)
meta.set_params(n_estimators=1000)
with pytest.raises(
ValueError,
match=r"n_estimators \(1000\) must not exceed number "
r"of base learners \(2\)"):
meta.fit(whas500.x, whas500.y)
def test_min_correlation(self):
base_estimators = [('gbm',
ComponentwiseGradientBoostingSurvivalAnalysis()),
('svm', FastSurvivalSVM())]
meta = EnsembleSelection(base_estimators,
scorer=score_cindex,
min_correlation=1.2)
self.assertRaisesRegex(
ValueError, "min_correlation must be in \[-1; 1\], but was 1.2",
meta.fit, self.x, self.y)
meta.set_params(min_correlation=-2.1)
self.assertRaisesRegex(
ValueError, "min_correlation must be in \[-1; 1\], but was -2.1",
meta.fit, self.x, self.y)
meta.set_params(min_correlation=numpy.nan)
self.assertRaisesRegex(
ValueError, "min_correlation must be in \[-1; 1\], but was nan",
meta.fit, self.x, self.y)
def test_correlation(make_whas500):
whas500 = make_whas500(with_mean=False, with_std=False, to_numeric=True)
base_estimators = [('gbm', ComponentwiseGradientBoostingSurvivalAnalysis()),
('svm', FastSurvivalSVM())]
meta = EnsembleSelection(base_estimators, scorer=score_cindex, correlation=None)
with pytest.raises(ValueError,
match="correlation must be one of 'pearson', 'kendall', "
"and 'spearman', but got None"):
meta.fit(whas500.x, whas500.y)
meta = EnsembleSelection(base_estimators, scorer=score_cindex, correlation=2143)
with pytest.raises(ValueError,
match="correlation must be one of 'pearson', 'kendall', "
"and 'spearman', but got 2143"):
meta.fit(whas500.x, whas500.y)
meta = EnsembleSelection(base_estimators, scorer=score_cindex, correlation="clearly wrong")
with pytest.raises(ValueError,
match="correlation must be one of 'pearson', 'kendall', "
"and 'spearman', but got 'clearly wrong'"):
meta.fit(whas500.x, whas500.y)
def test_squared_loss(make_whas500):
whas500_data = make_whas500(with_std=False, to_numeric=True)
model = ComponentwiseGradientBoostingSurvivalAnalysis(loss="squared", n_estimators=100, random_state=0)
model.fit(whas500_data.x, whas500_data.y)
time_predicted = model.predict(whas500_data.x)
time_true = whas500_data.y["lenfol"]
event_true = whas500_data.y["fstat"]
rmse_all = numpy.sqrt(mean_squared_error(time_true, time_predicted))
assert round(abs(rmse_all - 793.6256945839657), 7) == 0
rmse_uncensored = numpy.sqrt(mean_squared_error(time_true[event_true], time_predicted[event_true]))
assert round(abs(rmse_uncensored - 542.83358120153525), 7) == 0
cindex = model.score(whas500_data.x, whas500_data.y)
assert round(abs(cindex - 0.7777082862), 7) == 0
def _create_ensemble(self, **kwargs):
boosting_grid = ParameterGrid({
"n_estimators": [100, 250],
"subsample": [1.0, 0.75, 0.5]
})
svm_grid = ParameterGrid({"alpha": 2.**numpy.arange(-9, 5, 2)})
base_estimators = []
for i, params in enumerate(boosting_grid):
model = ComponentwiseGradientBoostingSurvivalAnalysis(
random_state=0, **params)
base_estimators.append(("gbm_%d" % i, model))
for i, params in enumerate(svm_grid):
model = FastSurvivalSVM(max_iter=100, random_state=0, **params)
base_estimators.append(("svm_%d" % i, model))
cv = KFold(n_splits=4, shuffle=True, random_state=0)
meta = EnsembleSelection(base_estimators,
n_estimators=0.4,
scorer=score_cindex,
cv=cv,
**kwargs)
return meta
def test_fit_verbose(make_whas500):
whas500_data = make_whas500(with_std=False, to_numeric=True)
model = ComponentwiseGradientBoostingSurvivalAnalysis(n_estimators=10, verbose=1, random_state=0)
model.fit(whas500_data.x, whas500_data.y)
def test_fit_verbose(self):
model = ComponentwiseGradientBoostingSurvivalAnalysis(n_estimators=10, verbose=1, random_state=0)
model.fit(self.x, self.y)
def test_feature_importances(self):
model = ComponentwiseGradientBoostingSurvivalAnalysis(n_estimators=100, random_state=0)
model.fit(self.x, self.y)
self.assertEqual(self.x.shape[1] + 1, len(model.feature_importances_))
# data_y
# df.groupby('status').count()
# Part 2: Componentwise Gradient Boosting for Survival Analysis
from sklearn.model_selection import ShuffleSplit, GridSearchCV
from sksurv.column import encode_categorical
from sksurv.metrics import concordance_index_censored
from sksurv.ensemble import ComponentwiseGradientBoostingSurvivalAnalysis
import warnings
warnings.filterwarnings("ignore", category=UserWarning)
## create estimator
estimator = ComponentwiseGradientBoostingSurvivalAnalysis(loss='coxph',
random_state=0)
## define a function for evaluating the performance of models during grid search using Harrell's concordance index
def score_survival_model(model, X, y):
prediction = model.predict(X)
result = concordance_index_censored(y['status'], y['time_to_event'],
prediction)
return result[0]
param_grid = {
'learning_rate': [0.001, 0.01, 0.1, 0.5, 1],
'n_estimators': [100, 200, 500, 1000]
}