def _make_data_space(self):
"""Пространство поиска параметров данных модели"""
space = {
"freq": hyper.make_choice_space("freq", Freq),
"lags": hyper.make_choice_space("lags", lags()),
}
return space
def _make_data_space(self):
"""Пространство поиска параметров данных модели"""
space = {
'freq': hyper.make_choice_space('freq', Freq),
'lags': hyper.make_choice_space('lags', lags())
}
return space
def test_optimize_hyper(monkeypatch):
space = {
'data': {
'freq': hyper.make_choice_space('freq', Freq),
'lags': hyper.make_choice_space('lags_range', range(1, 4))
},
'model': {
'one_hot_max_size':
hyper.make_choice_space('one_hot_max_size', hyper.ONE_HOT_SIZE),
'learning_rate':
hyper.make_log_space('learning_rate', 0.1, 0.1),
'depth':
hyper.make_choice_space('depth', range(1, 9)),
'l2_leaf_reg':
hyper.make_log_space('l2_leaf_reg', 2.3, 0.3),
'random_strength':
hyper.make_log_space('rand_strength', 1.3, 0.3),
'bagging_temperature':
hyper.make_log_space('bagging_temperature', 1.4, 0.4)
}
}
params = {
'data': {
'freq': Freq.yearly,
'lags': 1
},
'model': {
'one_hot_max_size': 2,
'learning_rate': 0.1,
'depth': 6,
'l2_leaf_reg': 2.3,
'random_strength': 1.3,
'bagging_temperature': 1.4
}
}
monkeypatch.setattr(hyper, 'MAX_SEARCHES', 2)
monkeypatch.setattr(hyper, 'make_model_space', lambda x: space['model'])
date = '2018-09-03'
pos = ('CHMF', 'RTKMP', 'SNGSP', 'VSMO', 'LKOH')
result = hyper.optimize_hyper(params, pos, pd.Timestamp(date),
cases.learn_pool, space['data'])
assert isinstance(result, dict)
assert result['data'] == dict(freq=Freq.quarterly, lags=1)
assert len(result['model']) == 6
assert result['model']['bagging_temperature'] == pytest.approx(
1.0557058439636)
assert result['model']['depth'] == 1
assert result['model']['l2_leaf_reg'] == pytest.approx(2.417498137284288)
assert result['model']['learning_rate'] == pytest.approx(
0.10806709959509389)
assert result['model']['one_hot_max_size'] == 100
assert result['model']['random_strength'] == pytest.approx(
1.0813796592585887)
def _make_data_space(self):
"""Пространство поиска параметров данных модели"""
space = {
"ew_lags": hp.uniform("ew_lags", *ew_lags(self.PARAMS)),
"returns_lags": hyper.make_choice_space("returns_lags", returns_lags()),
}
return space
def _make_data_space(self):
"""Пространство поиска параметров данных модели"""
space = {
'ew_lags': hp.uniform('ew_lags', *ew_lags(self.PARAMS)),
'returns_lags': hyper.make_choice_space('returns_lags',
returns_lags())
}
return space
'freq': Freq.yearly,
'lags': 1
},
'model': {
'bagging_temperature': 1.3463876077482095,
'depth': 3,
'l2_leaf_reg': 1.8578444629373057,
'learning_rate': 0.09300426944876264,
'one_hot_max_size': 2,
'random_strength': 1.0464151963029267
}
}
SPACE = {
'data': {
'freq': hyper.make_choice_space('freq', Freq),
'lags_range': hyper.make_choice_space('lags_range', list(range(1, 4)))
},
'model': {
'one_hot_max_size':
hyper.make_choice_space('one_hot_max_size', hyper.ONE_HOT_SIZE),
'learning_rate':
hyper.make_log_space('learning_rate', 0.1, 0.1),
'depth':
hyper.make_choice_space('depth', list(range(1, 9))),
'l2_leaf_reg':
hyper.make_log_space('l2_leaf_reg', 2.3, 0.3),
'random_strength':
hyper.make_log_space('rand_strength', 1.3, 0.3),
'bagging_temperature':
hyper.make_log_space('bagging_temperature', 1.4, 0.4)