def fetch_predict_data(self, ref_date: str, alpha_model: ModelBase):
return fetch_predict_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,
fillna=True)
def predict(self, ref_date: str, x: pd.DataFrame = None) -> pd.DataFrame:
if x is None:
predict_data = fetch_predict_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 = predict_data['predict']['x']
codes = predict_data['predict']['code']
else:
x_values = x.values
codes = x.index
model = self._fetch_latest_model(ref_date)
return pd.DataFrame(model.predict(x_values).flatten(), index=codes)
def model_predict(self, ref_date: str) -> pd.DataFrame:
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['predict']['x'][ref_date]
settlement_data = self.cached_data['settlement']
codes = settlement_data.loc[settlement_data.trade_date == ref_date,
'code'].values
else:
data = fetch_predict_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['predict']['x']
codes = data['predict']['code']
prediction = self.model.predict(ne_x).flatten()
return pd.DataFrame({'prediction': prediction, 'code': codes})