def test_grid_search_tuner(tmpdir):
train_set = pd.DataFrame({
'id': ["id1", "id2", "id3", "id3"],
'date':
pd.to_datetime(
["2016-01-01", "2016-02-01", "2016-03-01", "2016-04-01"]),
'x': [.2, .9, .3, .3],
'target': [0, 1, 0, 1]
})
eval_fn = roc_auc_evaluator(target_column="target")
space = {
'learning_rate': lambda: [1e-3, 1e-2, 1e-1],
'num_estimators': lambda: [1, 2],
'silent': lambda: [True]
}
@curry
def param_train_fn(space, train_set):
return xgb_classification_learner(
features=["x"],
target="target",
learning_rate=space["learning_rate"],
num_estimators=space["num_estimators"])(train_set)
split_fn = out_of_time_and_space_splitter(n_splits=2,
in_time_limit="2016-05-01",
space_column="id",
time_column="date")
tuning_log = grid_search_cv(space=space,
train_set=train_set,
param_train_fn=param_train_fn,
split_fn=split_fn,
eval_fn=eval_fn)
assert len(tuning_log) == 3 * 2
space = {
'learning_rate': lambda: [1e-3, 1e-2, 1e-1, 1],
'num_estimators': lambda: [1, 2],
'silent': lambda: [True]
}
tuning_log = grid_search_cv(space=space,
train_set=train_set,
param_train_fn=param_train_fn,
split_fn=split_fn,
eval_fn=eval_fn)
assert len(tuning_log) == 4 * 2
def test_roc_auc_evaluator():
predictions = pd.DataFrame({
'target': [0, 1, 0, 1],
'prediction': [.2, .9, .3, .3]
})
eval_fn = roc_auc_evaluator(prediction_column="prediction",
target_column="target",
eval_name="eval_name")
result = eval_fn(predictions)
assert result["eval_name"] == 0.875