def factor_uqer_one_day(factor_name,
ref_date,
use_only_index_components=False,
risk_neutral=True):
risk_factors = ','.join(
['risk_factor_300.' + name for name in risk_factors_300])
sql = "select factor_uqer.Code, factor_uqer.{factor_name}, factor_data.申万一级行业, {risk_factors}," \
" return_300.D1LogReturn, return_300.isTradable from factor_uqer".format(factor_name=factor_name,
risk_factors=risk_factors)
sql += " join factor_data on factor_uqer.Date = factor_data.Date and factor_uqer.Code = factor_data.Code" \
" join risk_factor_300 on factor_uqer.Date = risk_factor_300.Date and factor_uqer.Code = risk_factor_300.Code" \
" join return_300 on factor_uqer.Date = return_300.Date and factor_uqer.Code = return_300.Code" \
" where factor_uqer.Date = '{ref_date}'".format(ref_date=ref_date)
df1 = pd.read_sql(sql, engine).dropna()
if df1.empty:
return None, None
df2 = pd.read_sql(
"select Code, 300Weight / 100. as benchmark from index_components "
"where Date ='{ref_date}'".format(ref_date=ref_date), engine)
df = pd.merge(df1, df2, on=['Code'], how='left').fillna(0.)
if use_only_index_components:
df = df[df.benchmark != 0.]
factors = df[['Code', factor_name]].set_index('Code')[factor_name]
industry = df['申万一级行业'].values
d1returns = df['D1LogReturn'].values
is_tradable = df['isTradable'].values
benchmark = df['benchmark'].values
risk_exp = df[risk_factors_300].values
risk_exp = np.concatenate([risk_exp, np.ones((len(risk_exp), 1))], axis=1)
if risk_neutral:
weights, analysis = factor_analysis(factors=factors,
industry=industry,
d1returns=d1returns,
detail_analysis=True,
benchmark=benchmark,
risk_exp=risk_exp,
is_tradable=is_tradable)
else:
weights, analysis = factor_analysis(factors=factors,
industry=industry,
d1returns=d1returns,
detail_analysis=True,
risk_exp=risk_exp,
is_tradable=is_tradable)
return weights, analysis
def test_factor_analysis_with_several_factors(self):
benchmark = np.random.randint(50, size=1000)
benchmark = benchmark / benchmark.sum()
industry = np.random.randint(30, size=1000)
factor_df = pd.DataFrame(np.random.randn(1000, 2), index=range(len(self.raw_factor)))
factor_weights = np.array([0.2, 0.8])
constraints = Constraints()
names = np.array(['a', 'b', 'c'])
constraints.add_exposure(names, self.risk_factor)
targets = self.risk_factor.T @ benchmark
for i, name in enumerate(names):
constraints.set_constraints(name, targets[i], targets[i])
weight_table, analysis_table = factor_analysis(factor_df,
factor_weights,
d1returns=self.d1returns,
industry=industry,
benchmark=benchmark,
risk_exp=self.risk_factor,
constraints=constraints)
weight = weight_table.weight
self.assertEqual(analysis_table['er'].sum() / analysis_table['er'].iloc[-1], 2.0)
np.testing.assert_array_almost_equal(weight @ self.risk_factor, benchmark @ self.risk_factor)
def test_factor_analysis(self):
benchmark = np.random.randint(50, size=1000)
benchmark = benchmark / benchmark.sum()
industry = np.random.randint(30, size=1000)
factor_series = pd.Series(self.raw_factor.flatten(),
index=range(len(self.raw_factor)))
weight_table, analysis_table = factor_analysis(
factor_series,
d1returns=self.d1returns,
industry=industry,
benchmark=benchmark,
risk_exp=self.risk_factor)
weight = weight_table.weight
self.assertEqual(analysis_table['er'].sum() / analysis_table['er'][-1],
2.0)
np.testing.assert_array_almost_equal(weight @ self.risk_factor,
benchmark @ self.risk_factor)
self.assertTrue(
weight @ factor_series.values > benchmark @ factor_series.values)