def Get_total_loading(ret, tech_data, rm, macro_data, finance_loading, dummy):
interval = np.arange(192)
# Technical indicators
tech_loading = tech.fb_reg_over_time(ret.ix[interval], tech_data,
interval).iloc[:, :-1]
# standardization
tech_temp = pd.DataFrame(scale(tech_loading, axis=0)[:-3, :])
# Market indicators
mkt_loading = widgets.rm_reg_ri(rm.ix[interval], ret.ix[interval])
# standardization
mkt_temp = pd.DataFrame(scale(mkt_loading, axis=0)[:-3, :])
# Macro indicators
macro_loading = macro.macro_reg_ret(ret.ix[interval],
macro_data.ix[interval])
# standardization
macro_temp = pd.DataFrame(scale(macro_loading, axis=0)[:-3, :])
# Financial factors
finance_loading_temp = finance_loading.iloc[:-3, :]
# standardization
finance_temp = pd.DataFrame(scale(finance_loading_temp, axis=0))
# Industry factor indicators
ind_loading = dummy
# Not standardized
ind_temp = pd.concat(
[ind_loading.iloc[:-3, :-2], ind_loading.iloc[:-3, -1]], axis=1)
# Stitching the data to get the total loading
total_loading = pd.concat(
[tech_temp, mkt_temp, macro_temp, finance_temp, ind_temp], axis=1
) # Initialize, and according to the study of the two generated data inside a large number of 0, so thrown
return total_loading
def Get_total_loading(ret,tech_data,rm,macro_data,finance_loading,dummy):
interval = np.arange(192)
#技术指标
tech_loading = tech.fb_reg_over_time(ret.ix[interval], tech_data,interval).iloc[:,:-1]
#标准化
tech_temp = pd.DataFrame(scale(tech_loading,axis=0)[:-3,:])
#市场指标
mkt_loading = widgets.rm_reg_ri(rm.ix[interval], ret.ix[interval])
#标准化
mkt_temp = pd.DataFrame(scale(mkt_loading,axis=0)[:-3,:])
#宏观指标
macro_loading = macro.macro_reg_ret(ret.ix[interval], macro_data.ix[interval])
#标准化
macro_temp = pd.DataFrame(scale(macro_loading,axis=0)[:-3,:])
#财务因子指标
finance_loading_temp = finance_loading.iloc[:-3,:]
#标准化
finance_temp = pd.DataFrame(scale(finance_loading_temp,axis=0))
#行业因子指标
ind_loading = dummy
#不标准化
ind_temp = pd.concat([ind_loading.iloc[:-3,:-2],ind_loading.iloc[:-3,-1]],axis=1)
#拼接数据，得到总的loading
total_loading = pd.concat([tech_temp,mkt_temp,macro_temp,finance_temp,ind_temp],axis=1) # 初始化，并且根据考察这两个生成的数据里面有大量0，所以扔了
return total_loading
'E:\\QuantProject2\\temp_data\\all_stocks_dummy.xlsx')
dummy[0] = dummy[0].apply(lambda x: x[:-3])
dummy = dummy.set_index([0])
dummy = dummy.ix[stkcd['stkcd'].values]
dummy = pd.DataFrame(dummy.values)
# 代码为000009的股票没有行业信息，将其dummy令为0
dummy.fillna(0, inplace=True)
# 用全部时间的数据来计算一个loading,然后用这个loading去回归收益率，通过显著月数来判断决定保留哪些技术指标
all_tech_loading = tech.fb_reg_over_all_time(ret, tech_data)
significant_days_tech = tech.ret_reg_loading(all_tech_loading, ret, dummy)
print("Significant_days_tech:")
print(significant_days_tech)
interval = np.arange(192)
significant_days_mkt = tech.ret_reg_loading(
widgets.rm_reg_ri(rm.ix[interval], ret.ix[interval]), ret, dummy)
print("Significant_days_market:")
print(significant_days_mkt)
significant_days_macro = tech.ret_reg_loading(
macro.macro_reg_ret(ret.ix[interval], macro_data.ix[interval]), ret,
dummy)
print("Significant_days_macro:")
print(significant_days_macro)
# 根据上面注释掉的那段程序的结果，删掉了EMA,trade这两个技术指标,以及最后两个宏观指标
tech_data.pop('EMA')
tech_data.pop('trade')
# 计算loading
loading = dict()
for i in range(73):
interval = range(i, i + 119)
#技术指标（删去了最后一个市值自变量，由于下面要用它来做WLS）
dummy=pd.read_excel('E:\\QuantProject2\\temp_data\\all_stocks_dummy.xlsx')
dummy[0]=dummy[0].apply(lambda x:x[:-3])
dummy=dummy.set_index([0])
dummy=dummy.ix[stkcd['stkcd'].values]
dummy=pd.DataFrame(dummy.values)
# Code for the 000009 stock no industry information, the dummy order to 0
dummy.fillna(0,inplace=True)
# Use the full time of the data to calculate a loading, and then use this loading to return to the rate of return,
# through a significant number of months to determine what the decision to retain the technical indicators
all_tech_loading = tech.fb_reg_over_all_time(ret, tech_data)
significant_days_tech=tech.ret_reg_loading(all_tech_loading,ret,dummy)
print ("Significant_days_tech:")
print (significant_days_tech)
interval=np.arange(192)
significant_days_mkt = tech.ret_reg_loading(widgets.rm_reg_ri(rm.ix[interval], ret.ix[interval]),ret,dummy)
print ("Significant_days_market:")
print (significant_days_mkt)
significant_days_macro = tech.ret_reg_loading(macro.macro_reg_ret(ret.ix[interval], macro_data.ix[interval]),ret,dummy)
print ("Significant_days_macro:")
print (significant_days_macro)
# Based on the results of the procedure noted above, the two technical indicators, EMA, trade,
# and the last two macro indicators
tech_data.pop('EMA')
tech_data.pop('trade')
# Calculation
loading=dict()
for i in range(73):
interval = range(i,i+119)
#Technical indicators (delete the last market value of the independent variables,
# because the following to use it to do WLS)
# 读入dummy 矩阵并选取其中我们所需的股票
dummy=pd.read_excel('E:\\QuantProject2\\temp_data\\all_stocks_dummy.xlsx')
dummy[0]=dummy[0].apply(lambda x:x[:-3])
dummy=dummy.set_index([0])
dummy=dummy.ix[stkcd['stkcd'].values]
dummy=pd.DataFrame(dummy.values)
# 代码为000009的股票没有行业信息，将其dummy令为0
dummy.fillna(0,inplace=True)
# 用全部时间的数据来计算一个loading,然后用这个loading去回归收益率，通过显著月数来判断决定保留哪些技术指标
all_tech_loading = tech.fb_reg_over_all_time(ret, tech_data)
significant_days_tech=tech.ret_reg_loading(all_tech_loading,ret,dummy)
print ("Significant_days_tech:")
print (significant_days_tech)
interval=np.arange(192)
significant_days_mkt = tech.ret_reg_loading(widgets.rm_reg_ri(rm.ix[interval], ret.ix[interval]),ret,dummy)
print ("Significant_days_market:")
print (significant_days_mkt)
significant_days_macro = tech.ret_reg_loading(macro.macro_reg_ret(ret.ix[interval], macro_data.ix[interval]),ret,dummy)
print ("Significant_days_macro:")
print (significant_days_macro)
# 根据上面注释掉的那段程序的结果，删掉了EMA,trade这两个技术指标,以及最后两个宏观指标
tech_data.pop('EMA')
tech_data.pop('trade')
# 计算loading
loading=dict()
for i in range(73):
interval = range(i,i+119)
#技术指标（删去了最后一个市值自变量，由于下面要用它来做WLS）
tech_loading = tech.fb_reg_over_time(ret.ix[interval], tech_data,interval).iloc[:-3,:-1]
#tech_loading = tech_loading.drop([5],axis=0)