def add_industry_share_value(self):
"""
持仓数据增加字段
industry 行业,
share_value 股票的流通是指
:return:
"""
index_class_obj = GetIndexClass()
index_class = index_class_obj.get_index_class()
index_class_obj.get_zero_index_class()
share_data_obj = GetShare()
share_data = share_data_obj.get_share('float_a_share_value')
def cal_industry_share_value(x, index_class, share_data):
date_time = x.name
stock_code = x['instrument_exchange']
industry = index_class.loc[date_time, stock_code]
share_value = share_data.loc[date_time, stock_code]
if not isinstance(industry, str):
industry = 'other'
return industry, share_value
self.position_data_df['industry'], self.position_data_df['share_value'] = \
zip(*self.position_data_df.apply(cal_industry_share_value, args=(index_class, share_data,), axis=1))
def neutralize_method(self, method):
index_class_obj = GetIndexClass()
index_class_obj.get_index_class()
index_class_obj.get_zero_index_class()
share_data = pd.DataFrame({})
if NeutralizeMethod.MARKET_VALUE.value in method:
share_data_obj = GetShare()
share_data = share_data_obj.get_share('float_a_share_value')
def cal_resid(data, index_class_obj, share_data, method):
print('data.name', data.name)
# 删除一些 , 因子数据为NAN的个股
data = data.dropna()
index_class_in_date = pd.DataFrame({})
share_data_in_date = pd.DataFrame({})
if NeutralizeMethod.INDUSTRY.value in method:
index_class_in_date = index_class_obj.get_index_class_in_date(
data.name)
if NeutralizeMethod.MARKET_VALUE.value in method:
share_data_in_date = share_data.loc[data.name].dropna()
share_data_in_date = pd.DataFrame(
{'float_a_share_value': share_data_in_date})
# 行业中性与流通市值中性化取交集
neutralize_data = index_class_in_date.join(share_data_in_date,
how='outer').dropna()
# 因子数据的股票list与中性化数据的股票list,取交集
stock_code_list = list(
set(data.index).intersection(set(neutralize_data.index)))
# 因子数据取 有效股票列表数据，并排序
factor = data[stock_code_list].sort_index()
# 中性化数据取 有效股票列表数据，并排序
neutralize_data = neutralize_data.reindex(
stock_code_list).sort_index()
# 回归
neutralize_data = sm.add_constant(neutralize_data)
model = sm.OLS(factor, neutralize_data)
fit_result = model.fit()
# 残差作为中性化后的数据
return fit_result.resid
self.raw_data = self.raw_data.apply(cal_resid,
args=(
index_class_obj,
share_data,
method,
),
axis=1)
return self.raw_data
def cal_factor_return(self, method='float_value_inverse'):
"""
method = {‘float_value_inverse’, ‘float_value_square_root’}
:param method:
:return:
"""
index_class_obj = GetIndexClass()
index_class_obj.get_index_class()
index_class_obj.get_zero_index_class()
share_data_obj = GetShare()
share_data = share_data_obj.get_share('float_a_share_value')
index_list = self.factor.index
factor_return_daily = {}
factor_t_value_dict = {}
for index in range(self.factor.shape[0]):
stock_return = self.stock_return.iloc[index].dropna()
factor_data = self.factor.iloc[index].dropna()
stock_list = list(
set(stock_return.index).intersection(set(factor_data.index)))
stock_return = stock_return[stock_list].sort_index()
print(index_list[index])
index_class_in_date = index_class_obj.get_index_class_in_date(
index_list[index]).reindex(stock_list).sort_index()
share_data_in_date = share_data.loc[index_list[index]].reindex(
stock_list).dropna()
share_data_in_date = pd.DataFrame({
'float_a_share_value':
share_data_in_date[stock_list].sort_index()
})
factor_data = pd.DataFrame(
{self.factor_name: factor_data[stock_list].sort_index()})
x = sm.add_constant(
pd.concat(
[index_class_in_date, factor_data, share_data_in_date],
axis=1))
if stock_return.empty:
factor_return_daily[index_list[index]] = None
factor_t_value_dict[index_list[index]] = None
continue
wls_model = None
weights = None
if method == 'float_value_inverse':
weights = (1. / share_data_in_date['float_a_share_value'])
weights[np.isinf(weights)] = 0
wls_model = sm.WLS(stock_return, x, weights=weights)
elif method == 'float_value_square_root':
weights = share_data_in_date['float_a_share_value'].values**0.5
wls_model = sm.WLS(stock_return, x, weights=weights)
weights.name = index_list[index]
if wls_model is None:
factor_return_daily[index_list[index]] = None
factor_t_value_dict[index_list[index]] = None
continue
else:
results = wls_model.fit()
factor_return_daily[index_list[index]] = results.params[
self.factor_name]
factor_t_value_dict[index_list[index]] = results.tvalues[
self.factor_name]
self.factor_t_value = pd.Series(factor_t_value_dict)
self.factor_return_daily = pd.Series(factor_return_daily)
self.factor_return['cumsum'] = self.factor_return_daily.cumsum() + 1
self.factor_return['cumprod'] = (
self.factor_return_daily.add(1)).cumprod()
self.factor_return['daily'] = self.factor_return_daily