def test_imputer_1(self):
index = pd.MultiIndex.from_product([['2014-01-30', '2014-02-28'], ['001', '002', '003', '004']],
names=['trade_date', 'ticker'])
data1 = pd.DataFrame(index=index, data=[1.0, 3.0, 3.0, np.nan, 5.0, 5.0, 6.0, 8.0])
factor_test1 = Factor(data=data1, name='test1')
data2 = pd.DataFrame(index=index, data=[3.0, 2.0, 3.0, 7.0, 7.0, np.nan, 6.0, 6.0])
factor_test2 = Factor(data=data2, name='test2')
data3 = pd.DataFrame(index=index, data=[3.0, 3.0, np.nan, 5.0, 6.0, 7.0, 6.0, 6.0])
factor_test3 = Factor(data=data3, name='test3')
fc = FactorContainer('2014-01-30', '2014-02-28', [factor_test1, factor_test2, factor_test3])
index = pd.MultiIndex.from_product([[dt(2014, 1, 30), dt(2014, 2, 28)], ['001', '002', '003', '004']],
names=['trade_date', 'ticker'])
calculated = FactorImputer(numerical_strategy=NAStrategy.MOST_FREQ).fit_transform(fc)
expected = pd.DataFrame({'test1': [1.0, 3.0, 3.0, 3.0, 5.0, 5.0, 6.0, 8.0],
'test2': [3.0, 2.0, 3.0, 7.0, 7.0, 6.0, 6.0, 6.0],
'test3': [3.0, 3.0, 3.0, 5.0, 6.0, 7.0, 6.0, 6.0]},
index=index)
assert_frame_equal(calculated, expected)
calculated = FactorImputer(numerical_strategy=NAStrategy.MEDIAN).fit_transform(fc)
expected = pd.DataFrame({'test1': [1.0, 3.0, 3.0, 3.0, 5.0, 5.0, 6.0, 8.0],
'test2': [3.0, 2.0, 3.0, 7.0, 7.0, 6.0, 6.0, 6.0],
'test3': [3.0, 3.0, 3.0, 5.0, 6.0, 7.0, 6.0, 6.0]},
index=index)
assert_frame_equal(calculated, expected)
industry = pd.DataFrame(index=index, data=['a', 'a', 'a', 'a', 'a', 'a', np.nan, 'a'])
factor_industry = Factor(data=industry, name='industry', property_dict={'type': FactorType.INDUSTY_CODE})
fc.add_factor(factor=factor_industry)
calculated = FactorImputer(numerical_strategy=NAStrategy.MEDIAN,
categorical_strategy=NAStrategy.CUSTOM,
custom_value='other').fit_transform(fc)
calculated.sort_index(axis=1, inplace=True)
expected = pd.DataFrame({'test1': [1.0, 3.0, 3.0, 3.0, 5.0, 5.0, 6.0, 8.0],
'test2': [3.0, 2.0, 3.0, 7.0, 7.0, 6.0, 6.0, 6.0],
'test3': [3.0, 3.0, 3.0, 5.0, 6.0, 7.0, 6.0, 6.0],
'industry': ['a', 'a', 'a', 'a', 'a', 'a', 'other', 'a']},
index=index,
dtype=object)
assert_frame_equal(calculated, expected)
def test_imputer_2(self):
index = pd.MultiIndex.from_product(
[['2014-01-30', '2014-02-28'], ['001', '002', '003', '004']],
names=['trade_date', 'ticker'])
data = pd.DataFrame(index=index,
data=[1.0, 3.0, 3.0, np.nan, 5.0, 5.0, 6.0, 8.0])
factor_test = Factor(data=data, name='test1')
index = pd.MultiIndex.from_product(
[[dt(2014, 1, 30), dt(2014, 2, 28)], ['001', '002', '003', '004']],
names=['trade_date', 'ticker'])
fi = FactorImputer(numerical_strategy=NAStrategy.MEAN)
calculated = fi.fit_transform(factor_test)
expected = pd.DataFrame(
{'test1': [1.0, 3.0, 3.0, 2.33333333333, 5.0, 5.0, 6.0, 8.0]},
index=index)
assert_frame_equal(calculated, expected)
fi.set_out_container(True)
calculated = fi.fit_transform(factor_test)
expected = FactorContainer(start_date='2014-01-30',
end_date='2014-02-28')
factor = Factor(data=pd.DataFrame(
{'test1': [1.0, 3.0, 3.0, 2.33333333333, 5.0, 5.0, 6.0, 8.0]},
index=index),
name='test1')
expected.add_factor(factor)
assert (isinstance(calculated, FactorContainer))
self.assertEqual(calculated.property, expected.property)
assert_frame_equal(calculated.data, expected.data)
def test_imputer_2(self):
index = pd.MultiIndex.from_product([['2014-01-30', '2014-02-28'], ['001', '002', '003', '004']],
names=['trade_date', 'ticker'])
data = pd.DataFrame(index=index, data=[1.0, 3.0, 3.0, np.nan, 5.0, 5.0, 6.0, 8.0])
factor_test = Factor(data=data, name='test1')
index = pd.MultiIndex.from_product([[dt(2014, 1, 30), dt(2014, 2, 28)], ['001', '002', '003', '004']],
names=['trade_date', 'ticker'])
fi = FactorImputer(numerical_strategy=NAStrategy.MEAN)
calculated = fi.fit_transform(factor_test)
expected = pd.DataFrame({'test1': [1.0, 3.0, 3.0, 2.33333333333, 5.0, 5.0, 6.0, 8.0]}, index=index)
assert_frame_equal(calculated, expected)
fi.set_out_container(True)
calculated = fi.fit_transform(factor_test)
expected = FactorContainer(start_date='2014-01-30',
end_date='2014-02-28')
factor = Factor(data=pd.DataFrame({'test1': [1.0, 3.0, 3.0, 2.33333333333, 5.0, 5.0, 6.0, 8.0]}, index=index),
name='test1')
expected.add_factor(factor)
assert (isinstance(calculated, FactorContainer))
self.assertEqual(calculated.property, expected.property)
assert_frame_equal(calculated.data, expected.data)
def test_imputer_1(self):
index = pd.MultiIndex.from_product(
[['2014-01-30', '2014-02-28'], ['001', '002', '003', '004']],
names=['trade_date', 'ticker'])
data1 = pd.DataFrame(index=index,
data=[1.0, 3.0, 3.0, np.nan, 5.0, 5.0, 6.0, 8.0])
factor_test1 = Factor(data=data1, name='test1')
data2 = pd.DataFrame(index=index,
data=[3.0, 2.0, 3.0, 7.0, 7.0, np.nan, 6.0, 6.0])
factor_test2 = Factor(data=data2, name='test2')
data3 = pd.DataFrame(index=index,
data=[3.0, 3.0, np.nan, 5.0, 6.0, 7.0, 6.0, 6.0])
factor_test3 = Factor(data=data3, name='test3')
fc = FactorContainer('2014-01-30', '2014-02-28',
[factor_test1, factor_test2, factor_test3])
index = pd.MultiIndex.from_product(
[[dt(2014, 1, 30), dt(2014, 2, 28)], ['001', '002', '003', '004']],
names=['trade_date', 'ticker'])
calculated = FactorImputer(
numerical_strategy=NAStrategy.MOST_FREQ).fit_transform(fc)
expected = pd.DataFrame(
{
'test1': [1.0, 3.0, 3.0, 3.0, 5.0, 5.0, 6.0, 8.0],
'test2': [3.0, 2.0, 3.0, 7.0, 7.0, 6.0, 6.0, 6.0],
'test3': [3.0, 3.0, 3.0, 5.0, 6.0, 7.0, 6.0, 6.0]
},
index=index)
assert_frame_equal(calculated, expected)
calculated = FactorImputer(
numerical_strategy=NAStrategy.MEDIAN).fit_transform(fc)
expected = pd.DataFrame(
{
'test1': [1.0, 3.0, 3.0, 3.0, 5.0, 5.0, 6.0, 8.0],
'test2': [3.0, 2.0, 3.0, 7.0, 7.0, 6.0, 6.0, 6.0],
'test3': [3.0, 3.0, 3.0, 5.0, 6.0, 7.0, 6.0, 6.0]
},
index=index)
assert_frame_equal(calculated, expected)
industry = pd.DataFrame(
index=index, data=['a', 'a', 'a', 'a', 'a', 'a', np.nan, 'a'])
factor_industry = Factor(
data=industry,
name='industry',
property_dict={'type': FactorType.INDUSTY_CODE})
fc.add_factor(factor=factor_industry)
calculated = FactorImputer(numerical_strategy=NAStrategy.MEDIAN,
categorical_strategy=NAStrategy.CUSTOM,
custom_value='other').fit_transform(fc)
calculated.sort_index(axis=1, inplace=True)
expected = pd.DataFrame(
{
'test1': [1.0, 3.0, 3.0, 3.0, 5.0, 5.0, 6.0, 8.0],
'test2': [3.0, 2.0, 3.0, 7.0, 7.0, 6.0, 6.0, 6.0],
'test3': [3.0, 3.0, 3.0, 5.0, 6.0, 7.0, 6.0, 6.0],
'industry': ['a', 'a', 'a', 'a', 'a', 'a', 'other', 'a']
},
index=index,
dtype=object)
assert_frame_equal(calculated, expected)
data_return = load_factor_data_from_csv('return.csv')
# 将数据改成未来1月收益
data_return = fwd_return(data_return)
factor_return = Factor(data=data_return,
name='1_Fwd_Return',
property_dict={'type': FactorType.FWD_RETURN, 'norm_type': FactorNormType.Industry_Neutral})
# 创建FactorContainer实例，加载所有的因子信息
fc = FactorContainer(start_date='2014-01-01',
end_date='2014-03-01',
factors=[factor_mv, factor_pb, factor_return, factor_industry_code])
# pipeline
# 第一步，处理极个别N/A, 有中位数替换
step_1 = ('imputer', FactorImputer(numerical_strategy=NAStrategy.MEDIAN,
categorical_strategy=NAStrategy.CUSTOM,
custom_value='other'))
# 第二部，去极值化
step_2 = ('winsorize', FactorWinsorizer(quantile_range=(5, 95)))
# 第三步，标准化
step_3 = ('std', FactorStandardizer())
# 第四步，中性化
step_4 = ('neutralize', FactorNeutralizer())
# 第五步，求因子IC
step_5 = ('ic', FactorIC())
pipeline = AlphaPipeline([step_1, step_2, step_3, step_4, step_5])
ic = pipeline.fit_predict(fc)
factor_return = Factor(data=data_return, name='1_Fwd_Return', property_dict={'type': FactorType.FWD_RETURN})
# 加载行业数据(早年的wind的行业代码不太全，可能用其他数据源的数据更好，此处仅做示例用)
data_industry_code = load_factor_data_from_csv('sw.csv')
factor_industry_code = Factor(data=data_industry_code,
name='industry_code',
property_dict={'type': FactorType.INDUSTY_CODE})
# 创建FactorContainer实例，加载所有的因子信息
fc = FactorContainer(start_date='2014-01-01',
end_date='2014-03-01',
factors=[factor_mv, factor_pb, factor_return, factor_industry_code])
# 第一步，处理极个别N/A, 有中位数替换
fc = FactorImputer(numerical_strategy=NAStrategy.MEDIAN,
categorical_strategy=NAStrategy.CUSTOM,
custom_value='other',
out_container=True).fit_transform(fc)
# 第二部，去极值化
fc = FactorWinsorizer(quantile_range=(5, 95),
out_container=True).fit_transform(fc)
# 第三步，标准化
fc = FactorStandardizer(out_container=True).fit_transform(fc)
# 第四步，中性化
fc = FactorNeutralizer(out_container=True).fit_transform(fc)
# 第五步，求因子IC
ic = FactorIC().predict(fc)
print (ic)