def _create_bankdata_experiment(predefined_kwargs, maker=None, need_test=False, user_kwargs=None):
target = 'y'
df = dsutils.load_bank().head(2000)
df[target] = LabelEncoder().fit_transform(df[target])
df_train, df_test = train_test_split(df, test_size=0.3, random_state=9527)
def maker_(*args, **kwargs):
return make_experiment(PlainModel, *args, **kwargs)
default_kwargs = dict(log_level='info')
predefined_kwargs.update(default_kwargs)
if maker is None:
maker = maker_
predefined_kwargs['search_space'] = PlainSearchSpace(enable_lr=True,
enable_nn=False, enable_dt=False, enable_dtr=False)
predefined_kwargs['hyper_model_options'] = {'transformer': MultiLabelEncoder}
if need_test:
predefined_kwargs['test_data'] = df_test
predefined_kwargs.update(user_kwargs)
return maker(df_train, target=target, task=const.TASK_BINARY, **predefined_kwargs)
def test_experiment_with_blood_full_features():
df = dsutils.load_blood()
target = 'Class'
df_train, df_test = train_test_split(df, train_size=0.8, random_state=335)
df_test.pop(target)
experiment = make_experiment(
PlainModel,
df,
target=target,
search_space=PlainSearchSpace(),
test_data=df_test,
feature_generation=True,
collinearity_detection=True,
drift_detection=True,
feature_selection=True,
down_sample_search=True,
down_sample_search_size=0.2,
feature_reselection=True,
pseudo_labeling=True,
random_state=335,
early_stopping_time_limit=1200,
# log_level='info',
)
estimator = experiment.run(max_trials=3)
print(estimator)
assert estimator is not None
step_names = [step[0] for step in estimator.steps]
assert step_names == [
StepNames.DATA_CLEAN, StepNames.MULITICOLLINEARITY_DETECTION,
'estimator'
]
def _create_experiment(predefined_kwargs, maker=None, need_test=False, user_kwargs=None):
df = dsutils.load_boston()
df['Constant'] = [0 for i in range(df.shape[0])]
df['Id'] = [i for i in range(df.shape[0])]
target = 'target'
df_train, df_test = train_test_split(df, test_size=0.2, random_state=1234)
df_test.pop(target)
df_train['Drifted'] = np.random.random(df_train.shape[0])
df_test['Drifted'] = np.random.random(df_test.shape[0]) * 100
def maker_(*args, **kwargs):
if 'random_state' not in kwargs.keys():
kwargs['random_state'] = 1234
return make_experiment(PlainModel, *args, **kwargs)
default_kwargs = dict(
log_level='info',
)
predefined_kwargs.update(default_kwargs)
if maker is None:
maker = maker_
predefined_kwargs['search_space'] = PlainSearchSpace(enable_lr=False,
enable_nn=False, enable_dt=False, enable_dtr=True)
if need_test:
predefined_kwargs['test_data'] = df_test
predefined_kwargs.update(user_kwargs)
return maker(df_train, target=target, task=const.TASK_REGRESSION, **predefined_kwargs)
def test_creator(self):
test_data = dsutils.load_blood()
eval_data = dsutils.load_blood()
make_options = {
'test_data': test_data,
'eval_data': eval_data,
"task": const.TASK_BINARY,
'target': "Class",
'feature_selection': True,
'feature_selection_strategy': "threshold",
"evaluation_metrics": "auto",
"evaluation_persist_prediction": True,
"report_render": 'excel',
"search_space": PlainSearchSpace(),
}
job_working_dir = common_util.get_temp_dir_path(
prefix="hyn_job_creator_test_")
exp = BloodDatasetJobEngine().create_experiment_with_params(
make_options, job_working_dir)
assert exp
assert isinstance(exp, CompeteExperiment)
run_options = {"max_trials": 2}
exp.run(**run_options)
assert (Path(job_working_dir) / "report.xlsx").exists()
def test_experiment_with_blood_simple():
df = dsutils.load_blood()
experiment = make_experiment(PlainModel,
df,
target='Class',
search_space=PlainSearchSpace())
estimator = experiment.run(max_trials=3)
print(estimator)
assert estimator is not None
def create_hyper_model(reward_metric='auc', optimize_direction='max'):
search_space = PlainSearchSpace()
searcher = make_searcher('random',
search_space_fn=search_space,
optimize_direction=optimize_direction)
hyper_model = PlainModel(searcher=searcher,
reward_metric=reward_metric,
callbacks=[])
return hyper_model
def create_plain_model(reward_metric='auc', optimize_direction='max',
with_encoder=False, with_dask=False):
search_space = PlainSearchSpace(enable_dt=True, enable_lr=True, enable_nn=False)
searcher = make_searcher('random', search_space_fn=search_space, optimize_direction=optimize_direction)
encoder = MultiLabelEncoder if with_encoder else None
cls = DaskPlainModel if with_dask else PlainModel
hyper_model = cls(searcher=searcher, reward_metric=reward_metric, callbacks=[SummaryCallback()],
transformer=encoder)
return hyper_model
def maker(df_train, target, df_eval, file_path):
experiment = make_experiment(
PlainModel,
df_train,
target=target,
test_data=df_eval.copy(),
drift_detection_threshold=0.4,
drift_detection_min_features=3,
drift_detection_remove_size=0.5,
search_space=PlainSearchSpace(enable_lr=False, enable_nn=False),
report_render='excel',
report_render_options={'file_path': file_path})
return experiment
def maker(df_train, target, df_eval, file_path):
from hypernets.experiment.report import ExcelReportRender
experiment = make_experiment(
PlainModel,
df_train,
target=target,
eval_data=df_eval,
test_data=df_eval.copy(),
drift_detection_threshold=0.4,
drift_detection_min_features=3,
drift_detection_remove_size=0.5,
search_space=PlainSearchSpace(enable_lr=False, enable_nn=False),
report_render=ExcelReportRender(file_path))
return experiment
def test_regression_task_report():
df = dsutils.load_boston()
df['Constant'] = [0 for i in range(df.shape[0])]
df['Id'] = [i for i in range(df.shape[0])]
target = 'target'
df_train, df_eval = train_test_split(df, test_size=0.2)
df_train['Drifted'] = np.random.random(df_train.shape[0])
df_eval['Drifted'] = np.random.random(df_eval.shape[0]) * 100
file_path = common_util.get_temp_file_path(prefix="report_excel_",
suffix=".xlsx")
print(file_path)
experiment = make_experiment(
PlainModel,
df_train,
target=target,
eval_data=df_eval.copy(),
test_data=df_eval.copy(),
drift_detection_threshold=0.4,
drift_detection_min_features=3,
drift_detection_remove_size=0.5,
search_space=PlainSearchSpace(enable_lr=False,
enable_nn=False,
enable_dt=False,
enable_dtr=True),
report_render='excel',
report_render_options={'file_path': file_path})
estimator = experiment.run(max_trials=3)
assert estimator is not None
mlr_callback = None
mle_callback = None
for callback in experiment.callbacks:
if isinstance(callback, MLReportCallback):
mlr_callback = callback
if isinstance(callback, MLEvaluateCallback):
mle_callback = callback
assert mlr_callback is not None
_experiment_meta: ExperimentMeta = mlr_callback.experiment_meta_
assert len(_experiment_meta.resource_usage) > 0
assert len(_experiment_meta.steps) == 5
assert os.path.exists(file_path)
assert mle_callback is not None
assert _experiment_meta.evaluation_metric is not None
assert len(_experiment_meta.prediction_stats) == 1
assert len(_experiment_meta.datasets) == 3
def run_experiment(train_df, check_as_local=True, **kwargs):
experiment = make_experiment(PlainModel,
train_df,
search_space=PlainSearchSpace(),
**kwargs)
estimator = experiment.run()
print(experiment.random_state, estimator)
assert estimator is not None
if check_as_local:
assert hasattr(estimator, 'as_local')
local_estimator = estimator.as_local()
assert not hasattr(local_estimator, 'as_local')
def test_experiment_with_data_adaption():
df = dsutils.load_bank()
df = MultiLabelEncoder().fit_transform(df)
mem_usage = int(df.memory_usage().sum())
experiment = make_experiment(
PlainModel,
df,
target='y',
search_space=PlainSearchSpace(),
data_adaption_memory_limit=mem_usage // 2,
log_level='info',
)
estimator = experiment.run(max_trials=3)
assert estimator is not None
assert estimator.steps[0][0] == 'data_adaption'
def test_experiment_with_blood_down_sample():
df = dsutils.load_blood()
experiment = make_experiment(
PlainModel,
df,
target='Class',
search_space=PlainSearchSpace(),
down_sample_search=True,
down_sample_search_size=0.1,
down_sample_search_time_limit=300,
down_sample_search_max_trials=10,
# log_level='info',
)
estimator = experiment.run(max_trials=3)
print(estimator)
assert estimator is not None
def main():
df = dsutils.load_boston()
df_train, df_eval = train_test_split(df, test_size=0.2)
search_space = PlainSearchSpace(enable_lr=False,
enable_nn=False,
enable_dt=False,
enable_dtr=True)
experiment = make_experiment(PlainModel,
df_train,
target='target',
search_space=search_space,
report_render='excel')
estimator = experiment.run(max_trials=3)
print(estimator)
