def fetch_train_data(self, ref_date, alpha_model: ModelBase):
return fetch_train_phase(SqlEngine(self.data_source),
alpha_model.formulas,
ref_date,
self.freq,
self.universe,
self.batch,
self.neutralized_risk,
self.risk_model,
self.pre_process,
self.post_process,
self.warm_start,
fit_target=alpha_model.fit_target)
def fetch_train_data(self,
ref_date,
alpha_model: ModelBase):
return fetch_train_phase(self.engine,
alpha_model.formulas,
ref_date,
self.freq,
self.universe,
self.batch,
self.neutralized_risk,
self.risk_model,
self.pre_process,
self.post_process,
self.warm_start)
def train(self, ref_date: str):
train_data = fetch_train_phase(
self.data_meta.engine, self.data_meta.alpha_factors, ref_date,
self.data_meta.freq, self.data_meta.universe, self.data_meta.batch,
self.data_meta.neutralized_risk, self.data_meta.risk_model,
self.data_meta.pre_process, self.data_meta.post_process,
self.data_meta.warm_start)
x_values = train_data['train']['x']
y_values = train_data['train']['y']
self.alpha_model.fit(x_values, y_values)
self.models[ref_date] = copy.deepcopy(self.alpha_model)
self.is_updated = False
def model_train(self, ref_date: str):
if not self.is_const_model:
if self.cached_data and ref_date in self.scheduled_dates:
ref_date = dt.datetime.strptime(ref_date, '%Y-%m-%d')
ne_x = self.cached_data['train']['x'][ref_date]
ne_y = self.cached_data['train']['y'][ref_date]
else:
data = fetch_train_phase(self.data_source, self.features,
ref_date, self.freq, self.universe,
self.batch, self.neutralize_risk,
self.risk_model, self.pre_process,
self.post_process, self.warm_start)
ne_x = data['train']['x']
ne_y = data['train']['y']
self.model.fit(ne_x, ne_y)
