def check_alpha_factor_update_date(self):
""" 检查所有Alpha因子最后更新时间 """
factor_name_list = AlphaFactor().get_all_alpha_factor_name()
result = pd.DataFrame([],
columns=['开始日期', '结束日期'],
index=factor_name_list)
for i in range(0, len(factor_name_list)):
factor_name = factor_name_list[i]
try:
print("######### 检查更新日期 %s 数据 ############" % factor_name)
factor = AlphaFactor().get_alpha_factor_exposure(factor_name)
factor = factor.T.dropna(how='all').T
result.loc[factor_name, '开始日期'] = factor.columns[0]
result.loc[factor_name, '结束日期'] = factor.columns[-1]
result.loc[factor_name,
"最后一天有效数据个数"] = factor.iloc[:, -1].count()
result.loc[factor_name, "最后一天股票个数"] = len(factor.iloc[:, -1])
result.loc[factor_name,
"最后一天有效数据比率"] = factor.iloc[:, -1].count() / len(
factor.iloc[:, -1])
except Exception as e:
print(e)
result.loc[factor_name, '开始日期'] = ""
result.loc[factor_name, '结束日期'] = ""
result.loc[factor_name, "最后一天有效数据个数"] = ""
result.loc[factor_name, "最后一天股票个数"] = ""
result.loc[factor_name, "最后一天有效数据比率"] = ""
print("########### %s 检查更新数据 为空 ！！！###########" % factor_name)
out_file = os.path.join(self.data_path, "AlphaFactorUpdateDate.xlsx")
we = WriteExcel(out_file)
ws = we.add_worksheet("更新数据")
num_format_pd = pd.DataFrame([],
columns=result.columns,
index=['format'])
num_format_pd.loc['format', :] = '0'
num_format_pd.loc['format', ['最后一天有效数据比率']] = '0.00%'
we.write_pandas(result,
ws,
begin_row_number=0,
begin_col_number=1,
num_format_pd=num_format_pd,
color="blue",
fillna=True)
we.close()
def cal_cma_factor_pct(self):
""" 计算高资产增长率股票相对于低资产增长率股票的超额收益（分成三组）投资水平风险代表投资风险水平 """
name = "CMA"
bp = AlphaFactor().get_alpha_factor_exposure("alpha_raw_asset_yoy")
stock_pct = Stock().read_factor_h5("Pct_chg")
date_series = list(set(stock_pct.columns) & set(bp.columns))
date_series.sort()
result = pd.DataFrame([], index=date_series, columns=[name])
for i_date in range(1, len(date_series)):
date = date_series[i_date]
last_date = date_series[i_date - 1]
data_date = pd.concat([bp[last_date], stock_pct[date]], axis=1)
data_date = data_date.dropna()
data_date.columns = ['LastAssetYOY', 'Return']
data_date = data_date.sort_values(by=['LastAssetYOY'])
location = int(len(data_date) / 3)
low_assetyoy_stock_pct_mean = data_date.loc[data_date.index[0:location], 'Return'].mean()
data_date = data_date.sort_values(by=['LastAssetYOY'], ascending=False)
location = int(len(data_date) / 3)
high_assetyoy_stock_pct_mean = data_date.loc[data_date.index[0:location], 'Return'].mean()
result.loc[date, name] = high_assetyoy_stock_pct_mean - low_assetyoy_stock_pct_mean
result = result.dropna()
result['CumSumReturn'] = result[name].cumsum()
result.to_csv(os.path.join(self.data_path, 'factor_return', 'FactorReturn_%s.csv' % name))
def cal_mfc_holding_alpha_exposure_date(fund_name, factor_name_list, date):
""" 计算泰达持仓对alpha因子的暴露（满仓）"""
from quant.project.multi_factor.alpha_model.exposure.alpha_factor import AlphaFactor
date = Date().get_trade_date_offset(date, 0)
try:
holding_data = MfcGetData().get_fund_security(date)
holding_data = holding_data[["基金名称", "证券代码", "市值", '证券类别']]
holding_data = holding_data[holding_data["基金名称"] == fund_name]
holding_data = holding_data[holding_data['证券类别'] == "股票"]
holding_data.columns = ["FundName", "StockCode", "Weight", 'Type']
holding_data['Weight'] = holding_data['Weight'] / holding_data[
'Weight'].sum()
holding_data.StockCode = holding_data.StockCode.map(
CodeFormat().stock_code_add_postfix)
holding_data.index = holding_data.StockCode
weight = holding_data
exposure = pd.DataFrame()
for factor_name in factor_name_list:
alpha = AlphaFactor().get_alpha_factor_exposure(factor_name)
exposure_add = pd.DataFrame(alpha[date])
exposure_add.columns = [factor_name]
exposure = pd.concat([exposure, exposure_add], axis=1)
data = pd.concat([weight, exposure], axis=1)
data = data.sort_values(by=['Weight'], ascending=True)
data = data.dropna(subset=["Weight"])
res = pd.DataFrame([], columns=factor_name_list, index=[date])
if data['Weight'].sum() > 0.0:
for i_col in range(len(exposure.columns)):
factor_name = exposure.columns[i_col]
data_factor = data[["Weight", factor_name]]
data_factor = data_factor.dropna()
exposure_sum = (data_factor["Weight"] *
data_factor[factor_name]).sum()
res.loc[date,
factor_name] = exposure_sum / data['Weight'].sum()
print(" Calculate Mfcteda Fund %s Alpha Exposure at %s" %
(fund_name, date))
else:
print(
" Calculate Mfcteda Fund %s At %s of Weight Stock is Zero"
% (fund_name, date))
return res
except Exception as e:
print(" Calculate Mfcteda Fund %s Alpha Exposure at %s is Null " %
(fund_name, date))
res = pd.DataFrame([], columns=factor_name_list, index=[date])
return res
def ew_to_all_major_alpha(self, stock_pool_name, beg_date, end_date, period):
""" 等权合成因子 剔除ICIR表现不好的因子及因子值变化太快的因子 """
factor_list = AlphaFactor().get_all_alpha_factor_file()
major_factor_list = list(set(factor_list.index))
for i in range(len(major_factor_list)):
major_factor = major_factor_list[i]
factor_select = factor_list[factor_list.index == major_factor]
factor_name_list = list(factor_select["因子名"].values)
result = self.ew_to_major_alpha(factor_name_list, stock_pool_name, beg_date, end_date, period)
AlphaSplit().save_alpha_res_exposure(result, major_factor, stock_pool_name)
def ew_to_alpha(self, stock_pool_name):
""" 等权大类因子 """
major_factor_list = AlphaFactor().get_major_alpha_name()
result = pd.Panel()
for i in range(len(major_factor_list)):
major_factor = major_factor_list[i]
alpha = AlphaSplit().get_alpha_res_exposure(major_factor, stock_pool_name)
result = pd.concat([result, alpha])
result.items = major_factor_list
result_mean = result.mean(axis=0)
AlphaSplit().save_alpha_res_exposure(result_mean, "alpha", stock_pool_name)
def alpha_pool(self, beg_date, end_date, period, stock_pool_name):
""" 更新Alpha因子 在某个股票池 """
# AlphaSplit().split_alpha_all(beg_date, end_date, "D", stock_pool_name)
# factor_name_list = AlphaFactor().get_all_alpha_factor_name()
# AlphaSummary().cal_all_factor_return("20040101", end_date, factor_name_list, period, stock_pool_name, 1)
# AlphaSummary().cal_all_factor_summary("20040101", end_date, factor_name_list, period, stock_pool_name, 1)
# AlphaConcat().ew_to_all_major_alpha(stock_pool_name, beg_date, end_date, period)
factor_name_list = AlphaFactor().get_major_alpha_name()
AlphaSummary().cal_all_factor_return("20040101", end_date, factor_name_list, period, stock_pool_name, 1)
AlphaSummary().cal_all_factor_summary("20040101", end_date, factor_name_list, period, stock_pool_name, 1)
# AlphaConcat().ew_to_alpha(stock_pool_name)
AlphaSummary().cal_all_factor_return("20040101", end_date, ["alpha"], period, stock_pool_name, 1)
AlphaSummary().cal_all_factor_summary("20040101", end_date, ["alpha"], period, stock_pool_name, 1)
AlphaSummary().concat_summary(stock_pool_name)
def split_alpha_all(self,
beg_date,
end_date,
period="W",
stock_pool_name="AllChinaStockFilter",
force=1):
""" 拆分所有Alpha """
alpha_factor_list = AlphaFactor().get_all_alpha_factor_name()
for i in range(0, len(alpha_factor_list)):
alpha_name = alpha_factor_list[i]
hdf_res_path = os.path.join(self.data_path, stock_pool_name,
"res_alpha\hdf")
file = os.path.join(hdf_res_path, alpha_name + '.h5')
if (not os.path.exists(file)) or (force == 1):
self.split_alpha(beg_date, end_date, alpha_name, period,
stock_pool_name)
else:
print("%s Already Exist" % alpha_name)
def cal_factor_exposure(self, beg_date, end_date, index_code):
""" 计算指标数值 """
short_term = 5
long_term = 90
data = AlphaFactor().get_alpha_factor_exposure("alpha_raw_roe")
weight = Index().get_weight(index_code)
date_series = Date().get_trade_date_offset(beg_date, end_date)
date_series = set(date_series)
data = data.dropna()
data['DiffRatio'] = data['Diff'] / data['CLOSE']
data['RawTimer'] = data['DiffRatio']
data['Timer'] = data['RawTimer'].map(self.score_average_diff)
file = os.path.join(self.data_path, 'exposure',
'%s_%s.csv' % (self.factor_name, index_code))
data = data.dropna(how="all")
data.to_csv(file)
for i in range(1, len(concat_data.columns)):
col = concat_data.columns[i]
corr.loc[col,
'Corr'] = concat_data.iloc[:,
0].corr(concat_data.loc[:,
col])
return corr
if __name__ == "__main__":
# Data
##########################################################################################
from quant.stock.barra import Barra
from quant.project.multi_factor.alpha_model.exposure.alpha_factor import AlphaFactor
name = 'alpha_raw_ep'
date = "20171229"
data_pandas = AlphaFactor().get_alpha_factor_exposure(name)
factor_series = data_pandas[date]
neutral_frame = Barra().get_factor_exposure_date(
date, type_list=['STYLE', 'INDUSTRY'])
params, t_values, factor_res = FactorNeutral().factor_exposure_neutral(
factor_series, neutral_frame)
print(params)
print(factor_res)
print(t_values)
##########################################################################################
def split_alpha(self,
beg_date,
end_date,
factor_name,
period="W",
stock_pool_name="AllChinaStockFilter"):
""" 计算残差Alpha 回归风格因子和行业 计算在风格和行业上的暴露 """
alpha = AlphaFactor().get_standard_alpha_factor(factor_name)
date_series = Date().get_trade_date_series(beg_date,
end_date,
period=period)
barra_date_series = Barra().get_exposure_date_series()
date_series = list(
set(date_series) & set(alpha.columns) & set(barra_date_series))
date_series.sort()
res_alpha = pd.DataFrame()
exposure_risk = pd.DataFrame()
for i_date in range(len(date_series)):
date = date_series[i_date]
alpha_date = pd.DataFrame(alpha[date])
alpha_date.columns = ['Alpha']
alpha_date = alpha_date.dropna()
risk_exposure = Barra().get_factor_exposure_date(
date, type_list=['STYLE', 'INDUSTRY', "COUNTRY"])
stock_pool = Stock().get_invest_stock_pool(
date=date, stock_pool_name=stock_pool_name)
stock_pool = list(
set(stock_pool) & set(risk_exposure.index)
& set(alpha_date.index))
stock_pool.sort()
alpha_date = alpha_date.loc[stock_pool, "Alpha"]
risk_exposure = risk_exposure.loc[stock_pool, :]
concat_data = pd.concat([alpha_date, risk_exposure], axis=1)
concat_data = concat_data.dropna()
if len(concat_data) > self.min_stock_number:
factor_val = concat_data.iloc[:, 0]
neutral_val = concat_data.iloc[:, 1:]
print("Split %s Alpha Exposure At %s %s" %
(factor_name, date, stock_pool_name))
model = sm.OLS(factor_val.values, neutral_val.values)
regress = model.fit()
params = pd.DataFrame(regress.params,
index=neutral_val.columns,
columns=['param'])
factor_res = factor_val - regress.predict(neutral_val)
params = pd.DataFrame(params)
params.columns = [date]
res_alpha_date = pd.DataFrame(factor_res)
res_alpha_date.columns = [date]
exposure_risk = pd.concat([exposure_risk, params], axis=1)
res_alpha = pd.concat([res_alpha, res_alpha_date], axis=1)
else:
print("Split %s Alpha Exposure At %s %s is Null" %
(factor_name, date, stock_pool_name))
res_alpha = FactorPreProcess().remove_extreme_value_mad(res_alpha)
res_alpha = FactorPreProcess().standardization(res_alpha)
self.save_alpha_risk_exposure(exposure_risk, factor_name,
stock_pool_name)
self.save_alpha_res_exposure(res_alpha, factor_name, stock_pool_name)
