多因子组合

1. 读取相应数据
2. 获取每日IC的相关系数
3. 剔除相关系数高的因子
4. 计算最优权重组合因子
5. 用Alphalens查看绩效

1_读取相应数据

from jaqs.data.dataapi import DataApi
from jaqs.data import DataView
import numpy as np
from datetime import datetime
import pandas as pd
import warnings
import alphalens
warnings.filterwarnings("ignore")
dataview_folder = 'JAQS_Data/hs300'
dv = DataView()
dv.load_dataview(dataview_folder)

D:\Anaconda3\lib\site-packages\statsmodels\compat\pandas.py:56: FutureWarning: The pandas.core.datetools module is deprecated and will be removed in a future version. Please use the pandas.tseries module instead.
  from pandas.core import datetools


Dataview loaded successfully.

def change_columns_index(signal):
    new_names = {}
    for c in signal.columns:
        if c.endswith('SZ'):
            new_names[c] = c.replace('SZ', 'XSHE')
        elif c.endswith('SH'):
            new_names[c] = c.replace('SH', 'XSHG')
    signal = signal.rename_axis(new_names, axis=1)
    signal.index = pd.Index(map(lambda x: datetime.strptime(str(x),"%Y%m%d") , signal.index))
    return signal

all_factors = ['pb', 'roe', 'price_div_dps', 'ps_ttm', 'pe_ttm', 'roa']

origin_factors = {f: change_columns_index(dv.get_ts(f).loc[20150105:]) for f in all_factors}

from fxdayu_alphaman.factor.factor import Factor
from fxdayu_alphaman.factor.admin import Admin
f = Factor()

# 横截面标准化/去极值
PN_handle = pd.Panel({name: f.standardize(f.winsorize(frame)) for name, frame in origin_factors.items()})
# 加干扰
PN_disturbed = pd.Panel({name: f.get_disturbed_factor(frame) for name, frame in PN_handle.iteritems()})

prices = change_columns_index(dv.get_ts('close_adj'))

2_获取每日IC的相关系数

def cal_daily_ic(factor_df):
    factor_data = alphalens.utils.get_clean_factor_and_forward_returns(factor_df.stack(), prices, quantiles=5)
    return alphalens.performance.mean_information_coefficient(factor_data, by_time='D')
daily_ic = {key: cal_daily_ic(value) for key, value in PN_disturbed.iteritems()}
daily_panel = pd.Panel(daily_ic)

factor_corr = daily_panel.minor_xs(5).corr()

print(factor_corr)

                     pb    pe_ttm  price_div_dps    ps_ttm       roa       roe
pb             1.000000  0.799523       0.673749  0.901894  0.566400  0.181035
pe_ttm         0.799523  1.000000       0.858901  0.700401  0.140374 -0.305356
price_div_dps  0.673749  0.858901       1.000000  0.577473  0.055112 -0.334345
ps_ttm         0.901894  0.700401       0.577473  1.000000  0.529426  0.247359
roa            0.566400  0.140374       0.055112  0.529426  1.000000  0.860634
roe            0.181035 -0.305356      -0.334345  0.247359  0.860634  1.000000

3_剔除相关系数高的因子

ic = daily_panel.minor_xs(5).mean()

def compare(corr, targets):
    index = list(reversed(targets.index))
    length = len(index)
    for i in range(length):
        name = index[i]
        if available(corr, name, index[i+1:]):
            yield name

def available(corr, target, compares):
    for c in compares:
        if corr.loc[target, c] > 0.9 or (corr.loc[target, c] < -0.9 ):
            return False
    return True

small_df = pd.DataFrame({i: daily_panel.minor_xs(5)[i] for i in compare(factor_corr, ic.nsmallest(2))})

big_df = pd.DataFrame({i: daily_panel.minor_xs(5)[i] for i in compare(factor_corr, ic.nlargest(2))})

ic_df = pd.concat([big_df,small_df], axis=1)

print(big_df.head())

                 roa       roe
date                          
2015-01-05  0.362524  0.209361
2015-01-06  0.280116  0.170898
2015-01-07  0.341391  0.281306
2015-01-08  0.224819  0.340688
2015-01-09  0.223593  0.281711

4_计算最优权重组合因子

1. 导入Alphaman
2. 获取IC权重的DataFrame

from fxdayu_alphaman.factor.admin import Admin
f_admin = Admin()
ic_weight_df = f_admin.get_ic_weight_df(big_df, 10, rollback_period=20)

new_factors = {name: PN_disturbed[name].stack() for name in ic_weight_df.columns}

new_factor = f_admin.ic_cov_weighted_factor(new_factors, ic_weight_df=ic_weight_df)

factor = new_factor.multifactor_value

5_用Alphalens查看绩效

factor_data = alphalens.utils.get_clean_factor_and_forward_returns(factor, prices, quantiles=5)
mean_return_by_q, std_err_by_q = alphalens.performance.mean_return_by_quantile(factor_data, by_date=True)

import matplotlib.pyplot as plt
alphalens.plotting.plot_cumulative_returns_by_quantile(mean_return_by_q, 5)
plt.show()

ic = alphalens.performance.factor_information_coefficient(factor_data)
alphalens.plotting.plot_ic_hist(ic)
mean_monthly_ic = alphalens.performance.mean_information_coefficient(factor_data, by_time='M')
alphalens.plotting.plot_monthly_ic_heatmap(mean_monthly_ic)
plt.show()