added weights for linear regression model

alphamind/data/dbmodel/models.py

@@ -590,6 +590,7 @@ class DailyPortfolios(Base):
     industry = Column(String(50), nullable=False)
     benchmark_weight = Column(Float(53), nullable=False)
     is_tradable = Column(Boolean, nullable=False)
+    factor = Column(JSONB)
 
 
 class DailyReturn(Base):

alphamind/examples/combined_model_training.py

@@ -21,8 +21,8 @@ Back test parameter settings
 """
 
 start_date = '2012-01-01'
-end_date = '2012-11-15'
-benchmark_code = 300
+end_date = '2017-11-20'
+benchmark_code = 905
 universe_name = ['zz500', 'hs300']
 universe = Universe(universe_name, universe_name)
 frequency = '5b'
@@ -35,9 +35,9 @@ neutralize_risk = ['SIZE'] + industry_styles
 constraint_risk = ['SIZE'] + industry_styles
 size_risk_lower = 0
 size_risk_upper = 0
-turn_over_target_base = 0.25
+turn_over_target_base = 0.05
 weight_gaps = [0.01, 0.02, 0.03, 0.04]
-benchmark_total_lower = 0.8
+benchmark_total_lower = 1.
 benchmark_total_upper = 1.
 horizon = map_freq(frequency)
 hedging_ratio = 0.
@@ -101,7 +101,7 @@ for ref_date in ref_dates:
     alpha_logger.info('trade_date: {0} training finished'.format(ref_date))
 
 
-frequency = '5b'
+frequency = '1b'
 ref_dates = makeSchedule(start_date, end_date, frequency, 'china.sse')
 
 const_model_factor_data = engine.fetch_data_range(universe,
@@ -255,7 +255,7 @@ for weight_gap in weight_gaps:
 
         leverage = result.weight_x.abs().sum()
 
-        ret = (result.weight_x - hedging_ratio * result.weight_y * leverage / result.weight_y.sum()).values @ result.dx.values
+        ret = (result.weight_x - hedging_ratio * result.weight_y * leverage / result.weight_y.sum()).values @ (np.exp(result.dx.values) - 1.)
         rets.append(np.log(1. + ret))
         executor.set_current(executed_pos)
         turn_overs.append(turn_over)
@@ -269,7 +269,7 @@ for weight_gap in weight_gaps:
     # index return
     index_return = engine.fetch_dx_return_index_range(benchmark_code, start_date, end_date, horizon=horizon,
                                                       offset=1).set_index('trade_date')
-    ret_df['index'] = np.log(index_return['dx'] + 1.)
+    ret_df['index'] = index_return['dx']
 
     ret_df.loc[advanceDateByCalendar('china.sse', ref_dates[-1], frequency)] = 0.
     ret_df = ret_df.shift(1)

alphamind/examples/formula_expression.py

@@ -19,24 +19,23 @@ start = dt.datetime.now()
 
 universe_name = 'zz500'
 
-factor_name = 'ROIC'
-expression = LAST(factor_name)
+factor_name = 'PE'
+expression = 1. / LAST(factor_name)
 
-alpha_factor_name = 'alpha_factor'
+alpha_factor_name = '1/PE'
 alpha_factor = {alpha_factor_name: expression}
 
 # end of formula definition
 
 engine = SqlEngine('postgresql+psycopg2://postgres:A12345678!@10.63.6.220/alpha')
 universe = Universe('custom', [universe_name])
-benchmark_code = 905
 neutralize_risk = ['SIZE'] + industry_styles
-freq = '2w'
+freq = '5b'
 n_bins = 5
 horizon = map_freq(freq)
 
 start_date = '2012-01-01'
-end_date = '2017-11-03'
+end_date = '2017-11-20'
 
 dates = makeSchedule(start_date,
                      end_date,
@@ -45,17 +44,18 @@ dates = makeSchedule(start_date,
 
 factor_all_data = engine.fetch_data_range(universe,
                                           alpha_factor,
-                                          dates=dates,
-                                          benchmark=benchmark_code)['factor']
+                                          dates=dates)['factor']
 return_all_data = engine.fetch_dx_return_range(universe, dates=dates, horizon=horizon)
 
 factor_groups = factor_all_data.groupby('trade_date')
 return_groups = return_all_data.groupby('trade_date')
-final_res = np.zeros((len(dates), n_bins))
+final_res = np.zeros((len(factor_groups.groups), n_bins))
+
+index_dates = []
 
 for i, value in enumerate(factor_groups):
     date = value[0]
-    data = value[1][['code', alpha_factor_name, 'isOpen', 'weight'] + neutralize_risk]
+    data = value[1][['code', alpha_factor_name, 'isOpen'] + neutralize_risk]
     codes = data.code.tolist()
     ref_date = value[0].strftime('%Y-%m-%d')
     returns = return_groups.get_group(date)
@@ -63,7 +63,8 @@ for i, value in enumerate(factor_groups):
     total_data = pd.merge(data, returns, on=['code']).dropna()
     risk_exp = total_data[neutralize_risk].values.astype(float)
     dx_return = total_data.dx.values
-    benchmark = total_data.weight.values
+
+    index_dates.append(date)
 
     f_data = total_data[[alpha_factor_name]]
     try:
@@ -73,29 +74,28 @@ for i, value in enumerate(factor_groups):
                                post_process=[winsorize_normal, standardize])
         res = er_quantile_analysis(er,
                                    n_bins=n_bins,
-                                   dx_return=dx_return,
-                                   benchmark=benchmark)
+                                   dx_return=dx_return)
     except Exception as e:
         print(e)
         res = np.zeros(n_bins)
 
-    final_res[i] = res / benchmark.sum()
+    final_res[i] = res
 
-df = pd.DataFrame(final_res, index=dates)
+df = pd.DataFrame(final_res, index=index_dates)
 
 start_date = advanceDateByCalendar('china.sse', dates[0], '-1d')
 df.loc[start_date] = 0.
 df.sort_index(inplace=True)
 
 fig, axes = plt.subplots(1, 2, figsize=(18, 6))
-df = df.cumsum().plot(ax=axes[0], title='Quantile Analysis for {0}'.format(factor_name))
+df = df.cumsum().plot(ax=axes[0], title='Quantile Analysis for {0}'.format(alpha_factor_name))
 
 # =================================================================== #
 
-factor_name = 'ROE'
-expression = LAST(factor_name)
+factor_name = 'PE'
+expression = DIFF(1./LAST(factor_name))
 
-alpha_factor_name = 'alpha_factor'
+alpha_factor_name = '1/PE_1w_diff'
 alpha_factor = {alpha_factor_name: expression}
 
 dates = makeSchedule(start_date,
@@ -105,17 +105,18 @@ dates = makeSchedule(start_date,
 
 factor_all_data = engine.fetch_data_range(universe,
                                           alpha_factor,
-                                          dates=dates,
-                                          benchmark=benchmark_code)['factor']
+                                          dates=dates)['factor']
 return_all_data = engine.fetch_dx_return_range(universe, dates=dates, horizon=horizon)
 
 factor_groups = factor_all_data.groupby('trade_date')
 return_groups = return_all_data.groupby('trade_date')
-final_res = np.zeros((len(dates), n_bins))
+final_res = np.zeros((len(factor_groups.groups), n_bins))
+
+index_dates = []
 
 for i, value in enumerate(factor_groups):
     date = value[0]
-    data = value[1][['code', alpha_factor_name, 'isOpen', 'weight'] + neutralize_risk]
+    data = value[1][['code', alpha_factor_name, 'isOpen'] + neutralize_risk]
     codes = data.code.tolist()
     ref_date = value[0].strftime('%Y-%m-%d')
     returns = return_groups.get_group(date)
@@ -123,7 +124,8 @@ for i, value in enumerate(factor_groups):
     total_data = pd.merge(data, returns, on=['code']).dropna()
     risk_exp = total_data[neutralize_risk].values.astype(float)
     dx_return = total_data.dx.values
-    benchmark = total_data.weight.values
+
+    index_dates.append(date)
 
     f_data = total_data[[alpha_factor_name]]
     try:
@@ -133,21 +135,20 @@ for i, value in enumerate(factor_groups):
                                post_process=[winsorize_normal, standardize])
         res = er_quantile_analysis(er,
                                    n_bins=n_bins,
-                                   dx_return=dx_return,
-                                   benchmark=benchmark)
+                                   dx_return=dx_return)
     except Exception as e:
         print(e)
         res = np.zeros(n_bins)
 
-    final_res[i] = res / benchmark.sum()
+    final_res[i] = res
 
-df = pd.DataFrame(final_res, index=dates)
+df = pd.DataFrame(final_res, index=index_dates)
 
 start_date = advanceDateByCalendar('china.sse', dates[0], '-1d')
 df.loc[start_date] = 0.
 df.sort_index(inplace=True)
 
-df = df.cumsum().plot(ax=axes[1], title='Quantile Analysis for {0}'.format(factor_name))
+df = df.cumsum().plot(ax=axes[1], title='Quantile Analysis for {0}'.format(alpha_factor_name))
 
 plt.show()
 print(dt.datetime.now() - start)
\ No newline at end of file

alphamind/model/linearmodel.py

@@ -90,6 +90,10 @@ class LinearRegression(ModelBase):
         obj_layout.impl = pickle.loads(base64.decodebytes(encoded))
         return obj_layout
 
+    @property
+    def weights(self):
+        return self.impl.coef_.tolist()
+
 
 if __name__ == '__main__':