Appearance
使用示例
量价因子 范例 1
python
from rqfactor import *
from rqfactor.extension import *
from rqfactor.engine_v2 import *
import rqdatac
rqdatac.init()
VWAP = Factor('total_turnover') / Factor('volume')
f = (RANK((VWAP - Factor('close'))) / RANK((VWAP + Factor('close'))))量价因子 范例 2
python
from rqfactor import *
from rqfactor.extension import *
from rqfactor.engine_v2 import *
import rqdatac
import numpy as np
import pandas as pd
rqdatac.init()
#定义自定义因子
def buy_volume(order_book_ids,start_date,end_date):
return rqdatac.get_capital_flow(order_book_ids,start_date,end_date).buy_volume.unstack('order_book_id').reindex(columns=order_book_ids,index =pd.to_datetime(rqdatac.get_trading_dates(start_date,end_date)))
def sell_volume(order_book_ids,start_date,end_date):
return rqdatac.get_capital_flow(order_book_ids,start_date,end_date).sell_volume.unstack('order_book_id').reindex(columns=order_book_ids,index =pd.to_datetime(rqdatac.get_trading_dates(start_date,end_date)))
BUY_VOLUME = UserDefinedLeafFactor('BUY_VOLUME',buy_volume)
SELL_VOLUME = UserDefinedLeafFactor('SELL_VOLUME',sell_volume)
f = DELTA(MA(BUY_VOLUME-SELL_VOLUME,13)/IF(MA(ABS(BUY_VOLUME-SELL_VOLUME),13) !=0,MA(ABS(BUY_VOLUME-SELL_VOLUME),13),np.nan),3)
d1='20190101'
d2='20200101'
df = execute_factor(f,rqdatac.index_components('000300.XSHG', d1),d1,d2)
#实例化引擎
engine=FactorAnalysisEngine()
#构建管道，对因子进行预处理
engine.append(('neutralization', Neutralization(industry='citics_2019', style_factors=['size','beta','earnings_yield','growth','liquidity','leverage','book_to_price','residual_volatility','non_linear_size'])))
#构建管道，添加因子分析器
engine.append(('rank_ic_analysis', ICAnalysis(rank_ic=True, industry_classification='sws',max_decay=20)))
engine.append(('quantile', QuantileReturnAnalysis(quantile=5, benchmark=None)))
engine.append(('return',FactorReturnAnalysis()))
#调仓周期为1，3，5
result = engine.analysis(df, 'daily', ascending=True, periods=[1,3,5], keep_preprocess_result=True)- 查看因子 IC 分析结果
python
In[]:
result['rank_ic_analysis'].summary()
Out[]:
P_1 P_3 P_5
mean 0.002531 -0.005552 -0.012799
std 0.062977 0.064005 0.060862
positive 117.000000 118.000000 107.000000
negative 127.000000 126.000000 137.000000
significance 0.024590 0.040984 0.036885
sig_positive 0.012295 0.004098 0.008197
sig_negative 0.008197 0.024590 0.008197
t_stat 0.627707 -1.354970 -3.284878
p_value 0.530785 0.176685 0.001171
skew 0.223259 -0.401776 -0.088231
kurtosis 0.176964 0.414680 -0.099484
ir 0.040185 -0.086743 -0.210293- 绘制因子分组收益率结果
python
result['quantile'].show()
- 绘制因子收益率结果
python
result['return'].show()
月度因子（自然月）范例
财务因子比较偏向价值投资，一般持有期会相对较长，这里采用以自然月为收益区间验证财务因子
python
from rqfactor import *
from rqdatac import *
import pandas as pd
init()
#因子定义
f = RANK((Factor('net_profit_parent_company_ttm_0')/Factor('equity_parent_company_ttm_0')) /(Factor('net_profit_parent_company_ttm_1')/Factor('equity_parent_company_ttm_1')), 'first', True)
d1 = '20210101'
d2 = '20211201'
ids= index_components('000300.XSHG',d1)
#因子数据
df = execute_factor(f,ids,d1,d2)
df = df.resample('BM').last()
#合成每个自然月的收益率数据
returns=get_price_change_rate(ids, start_date=d1, end_date=d2, expect_df=True)+1
#为了使用resample能更便捷聚合收益率数据，这里直接将收益率数据index平移一个交易日
returns.index=get_trading_dates(get_previous_trading_date(returns.index[0]), get_previous_trading_date(returns.index[-1]), market='cn')
returns.columns.name=''
returns.index=pd.DatetimeIndex(returns.index)
#returns.index平移后，聚合的为当月第二个交易日到下一个月第一个交易日的收益率数据
returns=returns.resample('BM').prod()-1
returns.index=pd.DatetimeIndex(returns.index.date)
returns.columns.name=''
# 构建管道，并将因子值和收益率传入分析器中进行计算
engine = FactorAnalysisEngine()
# engine.append(('winzorization-mad', Winzorization(method='mad')))
engine.append(('rank_ic_analysis', ICAnalysis(rank_ic=True, industry_classification='sws',max_decay=10)))
engine.append(('QuantileReturnAnalysis', QuantileReturnAnalysis(quantile=10, benchmark='000300.XSHG')))
result = engine.analysis(df, returns, ascending=True, periods=1, keep_preprocess_result=True)- 查看 IC 分析结果及分组情况
python
In[]:
result['rank_ic_analysis'].summary()
Out[]:
P_1
mean -0.006272
std 0.106604
positive 7.000000
negative 5.000000
significance 0.333333
sig_positive 0.083333
sig_negative 0.083333
t_stat -0.203807
p_value 0.842226
skew 0.281141
kurtosis 0.418097
ir -0.058834
In[]:
result['QuantileReturnAnalysis'].quantile_detail
Out[]:
order_book_id 600183.XSHG 603658.XSHG 601319.XSHG 600637.XSHG 002179.XSHE \
datetime
2021-01-29 NaN NaN NaN NaN NaN
2021-02-26 q5 q3 q4 q6 q7
2021-03-31 q5 q3 q4 q6 q7
2021-04-30 q3 q3 q4 q6 q6
2021-05-31 q7 q2 q8 q4 q9
2021-06-30 q7 q2 q8 q4 q9
2021-07-30 q7 q2 q8 q4 q9
2021-08-31 q7 q2 q8 q4 q9
2021-09-30 q9 q2 q7 q2 q3
2021-10-29 q9 q2 q7 q2 q3
2021-11-30 q9 q3 q2 q3 q6
2021-12-31 q9 q3 q2 q3 q6
order_book_id 002050.XSHE 600660.XSHG 600176.XSHG 601877.XSHG 000157.XSHE \
datetime
2021-01-29 NaN NaN NaN NaN NaN
2021-02-26 q6 q3 q4 q5 q7
2021-03-31 q6 q3 q4 q5 q7
2021-04-30 q4 q8 q9 q5 q7
2021-05-31 q7 q9 q10 q2 q8
2021-06-30 q7 q9 q10 q2 q8
2021-07-30 q7 q9 q10 q2 q8
2021-08-31 q7 q9 q10 q2 q8
2021-09-30 q5 q8 q10 q5 q2
2021-10-29 q5 q8 q10 q5 q2
2021-11-30 q5 q5 q10 q3 q2
2021-12-31 q5 q5 q10 q3 q2
order_book_id ... 002311.XSHE 601398.XSHG 601111.XSHG 002384.XSHE \
datetime ...
2021-01-29 ... NaN NaN NaN NaN
2021-02-26 ... q6 q4 q10 q1
2021-03-31 ... q6 q4 q10 q1
2021-04-30 ... q6 q7 q10 q1
2021-05-31 ... q7 q4 q9 q2
2021-06-30 ... q7 q4 q9 q2
2021-07-30 ... q7 q4 q9 q2
2021-08-31 ... q2 q4 q9 q2
2021-09-30 ... q3 q6 q1 q3
2021-10-29 ... q3 q6 q1 q3
2021-11-30 ... q1 q6 q10 q8
2021-12-31 ... q1 q6 q10 q8
order_book_id 002594.XSHE 600690.XSHG 002600.XSHE 000661.XSHE 600019.XSHG \
datetime
2021-01-29 NaN NaN NaN NaN NaN
2021-02-26 q10 q8 q1 q7 q8
2021-03-31 q10 q8 q1 q7 q8
2021-04-30 q9 q9 q10 q5 q8
2021-05-31 q2 q7 q8 q6 q10
2021-06-30 q2 q7 q8 q6 q10
2021-07-30 q2 q7 q8 q6 q10
2021-08-31 q2 q7 q8 q6 q10
2021-09-30 q1 q8 q1 q5 q10
2021-10-29 q1 q8 q1 q5 q10
2021-11-30 q1 q2 q6 q5 q9
2021-12-31 q1 q2 q6 q5 q9
order_book_id 002008.XSHE
datetime
2021-01-29 NaN
2021-02-26 q9
2021-03-31 q9
2021-04-30 q2
2021-05-31 q9
2021-06-30 q9
2021-07-30 q9
2021-08-31 q9
2021-09-30 q6
2021-10-29 q6
2021-11-30 q9
2021-12-31 q9
[12 rows x 300 columns]- 绘制 IC 分析结果
python
result['rank_ic_analysis'].show()
F-Score 因子范例
F-Score 因子出自 Piotroski 的论文《Value Investing: The Use of Historical Financial Statement Information to Separate Winners from Losers》，主要从三个维度(profitability、financial leverage/liquidity、operating efficiency)来衡量公司的基本面状况。作者选用了九个好实施的指标，并且每个指标使用“好”或“坏”评判该公司在该指标上的表现。F-Score 因子是这九个指标的加总。
| 指标 | 打分方式 |
|---|---|
| 资产收益率 | 大于零为 1，否则为 0 |
| 资产收益率变化率 | 大于零为 1，否则为 0 |
| 经营活动产生的现金流比总资产 | 大于零为 1，否则为 0 |
| 应计收益率 | 大于零为 0，否则为 1 |
| 长期负债率变化 | 大于零为 0，否则为 1 |
| 流动比率变化 | 大于零为 0，否则为 1 |
| 股票是否增发 | 是为 0，否则为 1 |
| 毛利率变化 | 大于零为 1，否则为 0 |
| 资产周转率 | 大于零为 1，否则为 0 |
python
from rqfactor import *
from rqfactor.extension import *
from rqfactor.engine_v2 import *
import rqdatac as rq
import pandas as pd
rq.init()
import numpy as np
from rqfactor.engine_v2 import execute_factor
#自定义算子，使得因子值大于0时返回1，否则返回0
def fillter(df):
df[df>0]=1
df[df<=0]=0
df=df.fillna(0)
return df
def Fillter(f):
return UnaryCrossSectionalFactor(fillter, f)
#自定义算子，使得因子值大于0时返回0，否则返回1
def fillter_1(df):
df[df<0]=1
df[df>=0]=0
df=df.fillna(0)
return df
def Fillter_1(f):
return UnaryCrossSectionalFactor(fillter_1, f)
##构建因子
#1.ROA因子，大于0得1分，否则0分
f1 = Fillter(Factor('return_on_asset_ttm'))
#2.Δroa因子，大于0得1分，否则0分
f2 = Fillter(Factor('net_profit_mrq_0')/Factor('total_assets_mrq_0')-Factor('net_profit_mrq_3')/Factor('total_assets_mrq_3'))
#3.cfoa因子，大于0得1分，否则0分
f3 = Fillter(Factor('cash_flow_from_operating_activities_ttm_0')/Factor('total_assets_ttm_0'))
#4.应计利润,小于0得1分，否则0分
f4 = Fillter_1((Factor('profit_from_operation_ttm_0')-Factor('cash_flow_from_operating_activities_ttm_0'))/Factor('total_assets_ttm_0'))
#5.ΔLEVER,小于0得1分，否则0分,需对银行类别进行处理
def lever(order_book_ids, start_date, end_date):
trading_dates=rq.get_trading_dates(start_date, end_date, market='cn')
a=rq.get_factor(order_book_ids,'non_current_liabilities_mrq_0',start_date, end_date)['non_current_liabilities_mrq_0']
b=rq.get_factor(order_book_ids,'total_assets_mrq_0',start_date, end_date)['total_assets_mrq_0']
lever=a/b
lever=lever.unstack('order_book_id')
lever.columns.name=''
lever.index.name=''
lever=lever.reset_index(drop = False)
lever.index=lever[''].tolist()
lever.index=trading_dates
lever=lever.drop(columns='')
for key,value in lever.iteritems():
if key in rq.get_industry('银行', source='citics', date=end_date):
lever[key]=rq.get_factor(key,'deposits',start_date, end_date).unstack('order_book_id')['deposits']\
+rq.get_factor(key,'bond_payable',start_date, end_date).unstack('order_book_id')['bond_payable']\
+rq.get_factor(key,'borrowings_from_central_banks', start_date, end_date).unstack('order_book_id')['borrowings_from_central_banks']
lever.index = pd.DatetimeIndex(lever.index)
return lever
LEVER=UserDefinedLeafFactor('LEVER', lever)
f5=Fillter_1(LEVER-REF(LEVER, 252))
#6.ΔLIQUID,大于0得1分，否则0分
def liquid(order_book_ids, start_date, end_date):
trading_dates=rq.get_trading_dates(start_date, end_date, market='cn')
a=rq.get_factor(order_book_ids,'current_assets',start_date, end_date).fillna(method='pad')
b=rq.get_factor(order_book_ids,'current_liabilities',start_date, end_date).fillna(method='pad')
liquid=a['current_assets']/b['current_liabilities']
liquid=liquid.unstack('order_book_id')
liquid.columns.name=''
liquid.index.name=''
liquid=liquid.reset_index(drop = False)
liquid.index=liquid[''].tolist()
liquid.index=trading_dates
liquid=liquid.drop(columns='')
for key,value in liquid.iteritems():
bf_obi=rq.get_industry('银行', source='citics', date=end_date)+rq.get_industry('非银行金融', source='citics', date=end_date)
a=['cash_equivalent','deposits_of_interbank','precious_metals','lend_capital',
'financial_asset_held_for_trading','derivative_financial_assets','resale_financial_assets',
'interest_receivable','loans_advances_to_customers','financial_asset_available_for_sale'
'financial_asset_hold_to_maturity','loan_account_receivables']
b=['borrowings_from_central_banks','deposits_of_interbank','borrowings_capital','financial_liabilities',
'derivative_financial_liabilities','buy_back_security_proceeds','deposits','payroll_payable',
'tax_payable','dividend_payable']
if key in bf_obi:
liquid[key]=(rq.get_factor(key,a,start_date,end_date).sum(axis=1)/rq.get_factor(key,b,start_date,end_date).sum(axis=1)).unstack('order_book_id')
liquid.index=pd.DatetimeIndex(liquid.index)
return liquid
LIQUID=UserDefinedLeafFactor('LIQUID', liquid)
f6 = Fillter(LIQUID-REF(LIQUID,252))
#7.EQ_OFFER,过去一年是否增发或配售新股,没有增发得1分，否则为0
def eq_offer(order_book_ids, start_date, end_date):
trading_dates=rq.get_trading_dates(start_date, end_date, market='cn')
c=pd.DataFrame(columns=order_book_ids,index=trading_dates)
for i in trading_dates:
pp=rq.get_private_placement(order_book_ids, start_date=rq.get_previous_trading_date(i,252,market='cn'),
end_date=i, progress='complete',issue_type='private', market='cn')
for b in pp.index.get_level_values('order_book_id').values:
c.loc[i, b] = 0
for i in trading_dates:
pp=rq.get_private_placement(order_book_ids, start_date=rq.get_previous_trading_date(i,252,market='cn'),
end_date=i, progress='complete',issue_type='private', market='cn')
for b in pp.index.get_level_values('order_book_id').values:
c.loc[i, b] = 0
c=c.fillna(1)
c.index = pd.DatetimeIndex(c.index)
return c
f7 = UserDefinedLeafFactor('EQ_OFFER', eq_offer) #自定义因子
#8.ΔMARGIN因子，大于0得1分，否则0分
f8 = Fillter((Factor('operating_revenue_mrq_0')-Factor('total_expense_mrq_0'))/Factor('operating_revenue_mrq_0')\
-(Factor('operating_revenue_mrq_3')-Factor('total_expense_mrq_3'))/Factor('operating_revenue_mrq_3'))
#9.ΔTURN：最新报告期资产周转率-上年同期资产周转率,大于0得1分，否则0分
f9 =Fillter(Factor('operating_revenue_mrq_0') / Factor('total_assets_mrq_0')-Factor('operating_revenue_mrq_3') / Factor('total_assets_mrq_3'))
f=f1+f2+f3+f4+f5+f6+f7+f8+f9
#检验因子
#print(execute_factor(f, rq.index_components('000300.XSHG', '20180201'), '20180101', '20180201'))
df=execute_factor(f, rq.index_components('000300.XSHG','20150101'), '20150101', '20200101')
engine = FactorAnalysisEngine()
engine.append(('neutralization',Neutralization(industry='sws',style_factors=['size','beta','momentum','growth','book_to_price','residual_volatility','non_linear_size'])))
engine.append(('rank_ic_analysis', ICAnalysis(rank_ic=False, industry_classification='sws')))
engine.append(('QuantileReturnAnalysis', QuantileReturnAnalysis(quantile=3, benchmark='000300.XSHG')))
result = engine.analysis(df, 'daily', ascending=False, periods=22, keep_preprocess_result=True)
# 绘制 IC 结果图
result['rank_ic_analysis'].show()
result['QuantileReturnAnalysis'].show()
