Barra Risk Model for China A market
===================================
This is a self-build project for studying purpose only. Any commercial
usage of this project is banned.
Dependencies
------------
- python 3.5
- pandas 0.23.0
- spyder 3.2.8
Sample
------
Download descriptors from Wind API and make factors
.. code:: bash
path='your_source_file'
#########################
## step1 wind update ##
#########################
from barra_risk_model.Wind_update.get_wind import update_wind
dt_range_dict={
# 'dt':(None,20181231)
'rf':(None,20181019)
,'totalA':(None,20181019)
,'descriptor_annually':None
,'index':(None,20181019)
,'descriptor_daily':(20181011,20181019)
,'descriptor_seasonal':20180930
}
update_wind(path,dt_range_dict)
#####################
## step1.5 growth ##
#####################
# only at the beginning of each year when new growth is downloaded
from barra_risk_model.Factor_manufacture.make_growth import make_growth
# example for update growth_r of 20111231
year_range=(2011,2012)
make_growth(year_range,path)
########################
## step2 integration ##
########################
from barra_risk_model.Factor_manufacture.descriptor_integration import descriptor_integration
dt_range=(20181008,20181019)
warm_start_quarterly=20170930 # warm start should be 1 seasons before for annual reporting and 1 seasons before for quaterly reporting
warm_start_yearly=20161231 # 2 years before
descriptor_integration(path,dt_range,warm_start_quarterly,warm_start_yearly)
########################
## step3 make factors ##
########################
from barra_risk_model.Factor_manufacture.make_factor import manufacture
dt_range_dict={
# 'exr':(20181008,20181019) #should cover one more day before to create return from price
# ,'srcap':(20181008,20181019)
# ,'ind':(20181008,20181019)
'styl':(20120101,20181019)
}
history_start_date=20120101 #525 before styl starting date
manufacture(path,dt_range_dict,history_start_date,mapping=True,ind_mapping_exists=True)
Make prediction of index volatility
.. code:: bash
#########################
## step4 factor return ##
#########################
from barra_risk_model.Factor_return.Factor_return_core import factor_return_manufacture
## start
#dt_range=(20120101,20181019)
#factor_return_manufacture(path,dt_range,pool='ZZ800',filter_stock_flag=True,filter_dt_flag=True)
# renewal (use more days to keep more stocks,suggest)
dt_range=(20180901,20181019)
factor_return_manufacture(path,dt_range,pool='ZZ800',filter_stock_flag=True,filter_dt_flag=False)
#######################################
## step5 factor cov and specfic risk ##
#######################################
import numpy as np
from RNWS import read,write
from barra_risk_model.Factor_matrix_core import cov_matrix_gen_range
from barra_risk_model.Specific_matrix_core import spc_matrix_gen_range
start=20140312
end=20181019
Fct_return= read.read_df(path+'/factor_return_data',file_pattern='Fct_return',start=start,end=end)
F_all= cov_matrix_gen_range(Fct_return)
Spc_return=read.read_df(path+'/factor_return_data',file_pattern='Spc_return',start=start,end=end)
SRCap=read.read_df(path+'/srcap','srcap',start=start,end=end)
stock_pool=read.read_srs(path+'/index/ZZ800','Stk_ZZ800',start=start,end=end)
X_all= read.read_dict(path+'/factor_return_data',file_pattern='X1',start=start,end=end)
S_all=spc_matrix_gen_range(Spc_return=Spc_return,X_all=X_all,stock_pool=stock_pool,cap=np.square(SRCap))
renewal_dt=20140312
F_all_tmp={dt:F_all[dt] for dt in read.reading_data.trading_dt[(read.reading_data.trading_dt>=renewal_dt)&(read.reading_data.trading_dt<=end)]}
S_all_tmp=S_all.loc[renewal_dt:]
write.write_dict(F_all_tmp,path=path+'/factor_cov_matrix',file_pattern='factor_cov')
write.write_df(S_all_tmp,path=path+'/specific_volatility',file_pattern='specofoc_vol')
#############################
# step6 one day prediction ##
#############################
import numpy as np
from RNWS import read,write
from barra_risk_model.Factor_matrix_core import cov_matrix_oneday
from barra_risk_model.Specific_matrix_core import spc_matrix_oneday
start=20170508
end=20180508
Fct_return= read.read_df(path+'/factor_return_data',file_pattern='Fct_return',start=start,end=end)
F_one=cov_matrix_oneday(Fct_return,dt=20180508)
Spc_return=read.read_df(path+'/factor_return_data',file_pattern='Spc_return',start=start,end=end)
X_all= read.read_dict(path+'/factor_return_data',file_pattern='X1',start=start,end=end)
stock_pool=read.read_srs(path+'/index/ZZ800','Stk_ZZ800',start=start,end=end)
SRCap=read.read_df(path+'/srcap','srcap',start=start,end=end)
S_one=spc_matrix_oneday(Spc_return=Spc_return,dt=20180508,X_all=X_all,stock_pool=stock_pool,cap=np.square(SRCap))
# S_one predict one day specific volatility, requiring 128 days to preoduce.
###############################
## step7 evaluate prediction ##
###############################
from barra_risk_model.Bias_stats import bs_F,bs,bs_window,bs_f_window,real_std,predicted_std
import datetime
import matplotlib.pyplot as plt
start=20140312
end=20181019
F_all= read.read_dict(path+'/factor_cov_matrix',file_pattern='factor_cov',start=start,end=end)
R_all= read.read_df(path+'/factor_return_data',file_pattern='R',start=start,end=end)
X_all= read.read_dict(path+'/factor_return_data',file_pattern='X1',start=start,end=end)
S_all= read.read_df(path+'/specific_volatility',file_pattern='specofoc_vol',start=start,end=end)
W= read.read_df(path+'/index/HS300',file_pattern='Stk_HS300',start=start,end=end,dat_col=3,inx_col=1)
W=W.replace('None',np.nan).astype(float)
W2=W.fillna(0)
#W2=W2.where(W2==0)
def plot_bs(R_all,F_all,X_all,S_all,W):
bs_f=bs_F(R_all,F_all,X_all,W)
bs_fw=bs_f_window(R_all,F_all,X_all,W)
bs_=bs(R_all,F_all,X_all,S_all,W)
bs_w=bs_window(R_all,F_all,X_all,S_all,W)
bs_fw.index=[datetime.datetime.strptime(str(i),'%Y%m%d') for i in bs_fw.index]
bs_w.index=[datetime.datetime.strptime(str(i),'%Y%m%d') for i in bs_w.index]
plt.figure(figsize=(10,6))
plt.plot(bs_w,label='bs')
plt.plot(bs_fw,label='bs_f')
plt.gcf().autofmt_xdate()
plt.legend()
plt.show()
print(bs_f,bs_)
def plot_vol(R_all,F_all,X_all,S_all,W):
rv=real_std(R_all,W)
pv=predicted_std(F_all,S_all,X_all,W)
rv.index=[datetime.datetime.strptime(str(i),'%Y%m%d') for i in rv.index]
pv.index=[datetime.datetime.strptime(str(i),'%Y%m%d') for i in pv.index]
plt.figure(figsize=(10,6))
plt.plot(rv,label='real volatility')
plt.plot(pv,label='predicted volatility')
plt.gcf().autofmt_xdate()
plt.legend()
plt.show()
plot_vol(R_all,F_all,X_all,S_all,W2)
rv=real_std(R_all,W2)
pv=predicted_std(F_all,S_all,X_all,W2)
pv2=pv.reset_index()
rv2=rv.reset_index()
pv3=pv.iloc[np.where((pv2['index']//100).pct_change()!=0)]
rv3=rv.iloc[np.where((rv2['index']//100).pct_change()!=0)]
rv3.index=[datetime.datetime.strptime(str(i),'%Y%m%d') for i in rv3.index]
pv3.index=[datetime.datetime.strptime(str(i),'%Y%m%d') for i in pv3.index]
plt.figure(figsize=(10,6))
plt.plot(rv3,label='real volatility')
plt.plot(pv3,label='predicted volatility')
plt.gcf().autofmt_xdate()
plt.legend()
plt.show()