def FundReturnQuarter(report_date):
"""
计算在给定时间点前后一个月 一个基金收益 利用基金复权净值计算
"""
T = 20
beg_date = Date().get_trade_date_offset(report_date, -T)
end_date = Date().get_trade_date_offset(report_date, T)
date_series = Date().get_trade_date_series(beg_date, end_date)
fund_code_list = Fund().get_fund_pool_code(date=report_date,
name="基金持仓基准基金池")
fund_code_list3 = Fund().get_fund_pool_code(date=report_date, name="量化基金")
fund_code_list2 = Fund().get_fund_pool_code(date="20180630", name="东方红基金")
fund_code_list.extend(fund_code_list2)
fund_code_list.extend(fund_code_list3)
fund_code_list = list(set(fund_code_list))
fund_code_list.sort()
fund_pct = Fund().get_fund_factor("Repair_Nav_Pct", date_series,
fund_code_list)
print(" Calculting fund Return At date %s " % report_date)
fund_pct[fund_pct == 0] = np.nan
fund_pct_period = pd.DataFrame(fund_pct.sum(skipna=True).values,
index=fund_pct.columns,
columns=[report_date])
fund_pct_period[fund_pct_period == 0] = np.nan
return fund_pct_period
def AllFundBarraDecomposeReturnQuarter(fund_holding_all, path, report_date):
"""
计算一个季报时间点 前后一个月 所有基金 拆分的 特异收益 风格收益 行业收益 和 市场收益
"""
# 参数
######################################################################################################
# report_date = '20171231'
# path = 'E:\\3_Data\\4_fund_data\\7_fund_select_stock\\'
# fund_holding_all = Fund().get_fund_holding_all()
# 持仓信息
######################################################################################################
fund_holding_date = fund_holding_all[fund_holding_all['Date'] ==
report_date]
# 基金池信息
######################################################################################################
fund_code_list = Fund().get_fund_pool_code(date=report_date,
name="基金持仓基准基金池")
fund_code_list3 = Fund().get_fund_pool_code(date=report_date, name="量化基金")
fund_code_list2 = Fund().get_fund_pool_code(date="20180630", name="东方红基金")
fund_code_list.extend(fund_code_list2)
fund_code_list.extend(fund_code_list3)
fund_code_list = list(set(fund_code_list))
fund_code_list.sort()
# cal fund alpha on style
######################################################################################################
for i in range(0, len(fund_code_list)):
fund_code = fund_code_list[i]
if i == 0:
result = FundBarraDecomposeReturnQuarter(fund_holding_date,
report_date, path,
fund_code)
else:
res = FundBarraDecomposeReturnQuarter(fund_holding_date,
report_date, path, fund_code)
result = pd.concat([result, res], axis=0)
######################################################################################################
file = "FundBarraDecomposeReturnQuarter" + report_date + '.csv'
file = os.path.join(path, "FundBarraDecomposeReturnQuarter", file)
result.to_csv(file)
def opt_date(self, end_date, end_last_date, turnover_control=False):
"""
单期优化
1、风格不能偏离整体中位数太多
2、仓位不能偏离整体中位数太多
3、单个基金上限
4、换手率约束
5、每个基金管理公司最多有两只（未约束）
6、基金个数约束（未约束）
"""
print("Project Nice Stock Fund Optimization Fund Weight at %s ......" % end_date)
# Fund Pool
path = os.path.join(self.data_path, 'fund_factor')
file = os.path.join(path, 'FundFactor_%s.csv' % end_date)
data = pd.read_csv(file, index_col=[0], encoding='gbk')
fund_benchmark_list = Fund().get_fund_pool_code("20181231", self.fund_pool_name)
fund_benchmark_list = list(set(fund_benchmark_list) & set(data.index))
fund_benchmark_list.sort()
data = data.loc[fund_benchmark_list, :]
# Name of Columns
barra_style_list = Barra().get_factor_name(type_list=['STYLE'])['NAME_EN'].values
position_list = Barra().get_factor_name(type_list=['COUNTRY'])['NAME_EN'].values
risk_factor_list = Barra().get_factor_name(type_list=['COUNTRY', "STYLE"])['NAME_EN'].values
alpha_col = [self.alpha_column]
use_col = list(alpha_col)
use_col.extend(risk_factor_list)
# Fund Pool Filter
data = data.dropna(subset=use_col)
alpha_values = data[alpha_col].values
data['UpRatio'] = self.fund_up_ratio
weight_up_values = data['UpRatio'].values
# BenchMark
fund_benchmark_list = Fund().get_fund_pool_code("20181231", self.benchmark_name)
fund_benchmark_list = list(set(fund_benchmark_list) & set(data.index))
fund_benchmark_list.sort()
# Val
stock_style_values = data[barra_style_list].values
stock_position_values = data[position_list].values
bench_style_values = data.loc[fund_benchmark_list, barra_style_list].median().values
bench_position_values = data.loc[fund_benchmark_list, position_list].median().values
bench_style_up_values = bench_style_values.T + self.style_deviate
bench_style_low_values = bench_style_values.T - self.style_deviate
bench_position_up_values = bench_position_values.T + self.position_deviate
bench_position_low_values = bench_position_values.T - self.position_deviate
if len(data) == 0:
print("Project Nice Stock Fund Length of Fund is %s at %s ...... is Zero" % (len(data), end_date))
else:
print("Project Nice Stock Fund Length of Fund is %s at %s ......" % (len(data), end_date))
w = cvx.Variable(len(data))
if turnover_control:
try:
file = os.path.join(self.data_path, 'fund_opt', 'FundOpt_%s.csv' % end_last_date)
weight_last = pd.read_csv(file, index_col=[0], encoding='gbk')
weight_last = weight_last.dropna(subset=[self.alpha_column])
if len(weight_last) == 0:
weight_last = pd.Series(index=data.index)
weight_last = weight_last.fillna(0.0)
turnover = 2.00
else:
weight_last = weight_last.loc[data.index, "Weight"]
weight_last = weight_last.fillna(0.0)
turnover = self.turnover
except Exception as e:
print(e)
weight_last = pd.Series(index=data.index)
weight_last = weight_last.fillna(0.0)
turnover = 2.00
print(len(weight_last))
weight_last_values = weight_last.values
prob = cvx.Problem(cvx.Maximize(alpha_values.T * w),
[cvx.sum(w) == 1,
w >= 0,
w <= weight_up_values,
cvx.sum(cvx.abs(w - weight_last_values)) <= turnover,
stock_style_values.T * w <= bench_style_up_values,
stock_style_values.T * w >= bench_style_low_values,
stock_position_values.T * w <= bench_position_up_values,
stock_position_values.T * w >= bench_position_low_values,
])
prob.solve()
else:
prob = cvx.Problem(cvx.Maximize(alpha_values.T * w),
[cvx.sum(w) == 1,
w >= 0,
w <= weight_up_values,
stock_style_values.T * w <= bench_style_up_values,
stock_style_values.T * w >= bench_style_low_values,
stock_position_values.T * w <= bench_position_up_values,
stock_position_values.T * w >= bench_position_low_values,
])
prob.solve()
print("status:", prob.status)
print("optimal value", prob.value)
weight = pd.DataFrame(w.value, columns=['Weight'], index=data.index)
weight['Weight'] = weight['Weight'].map(lambda x: 0.0 if x < 0.02 else x)
weight['Weight'] /= weight['Weight'].sum()
result_risk_exposure = weight.T.dot(data[risk_factor_list])
result_risk_exposure = result_risk_exposure.T
result_risk_exposure.columns = ['ResultExposure']
up_risk_exposure = pd.DataFrame(data[risk_factor_list].median())
up_risk_exposure.columns = ["UpExposure"]
up_risk_exposure.loc[barra_style_list, "UpExposure"] += self.style_deviate
up_risk_exposure.loc[position_list, "UpExposure"] += self.position_deviate
low_risk_exposure = pd.DataFrame(data[risk_factor_list].median())
low_risk_exposure.columns = ["LowExposure"]
low_risk_exposure.loc[barra_style_list, "LowExposure"] -= self.style_deviate
low_risk_exposure.loc[position_list, "LowExposure"] -= self.position_deviate
bench_risk_exposure = pd.DataFrame(data[risk_factor_list].median())
bench_risk_exposure.columns = ["BenchExposure"]
exposure_result = pd.concat([bench_risk_exposure, up_risk_exposure,
low_risk_exposure, result_risk_exposure], axis=1)
data = pd.concat([data, weight], axis=1)
data = data[data['Weight'] > 0.0]
result = pd.concat([data, exposure_result.T], axis=0)
col = ["SecName", "InvestType", "Corp", "SetupDate", "Weight",
"RegressAlpha", "RegressAlphaIR"]
col.extend(risk_factor_list)
result = result[col]
path = os.path.join(self.data_path, 'fund_opt')
file = os.path.join(self.data_path, 'fund_opt', 'FundOpt_%s.csv' % end_date)
if not os.path.exists(path):
os.makedirs(path)
result.to_csv(file)
