@Channelchan
2018-04-18T08:18:19.000000Z
字数 20278
阅读 51371
参数优化器
目录
- 因子参数优化
- 事件参数优化
- 拓展：外部自定义带参算法优化
from jaqs_fxdayu.data import DataViewimport warningswarnings.filterwarnings("ignore")dataview_folder = './Factor'dv = DataView()dv.load_dataview(dataview_folder)
Dataview loaded successfully.
import numpy as npdef mask_index_member():df_index_member = dv.get_ts('index_member')mask_index_member = ~(df_index_member >0) #定义信号过滤条件-非指数成分return mask_index_memberdef limit_up_down():# 定义可买卖条件——未停牌、未涨跌停trade_status = dv.get_ts('trade_status')mask_sus = trade_status == u'停牌'# 涨停dv.add_formula('up_limit', '(close - Delay(close, 1)) / Delay(close, 1) > 0.095', is_quarterly=False, add_data=True)# 跌停dv.add_formula('down_limit', '(close - Delay(close, 1)) / Delay(close, 1) < -0.095', is_quarterly=False, add_data=True)can_enter = np.logical_and(dv.get_ts('up_limit') < 1, ~mask_sus) # 未涨停未停牌can_exit = np.logical_and(dv.get_ts('down_limit') < 1, ~mask_sus) # 未跌停未停牌return can_enter,can_exit
mask = mask_index_member()can_enter,can_exit = limit_up_down()
from jaqs_fxdayu.research import Optimizer
'''
:param dataview: 包含了计算公式所需要的所有数据的jaqs.data.DataView对象
:param formula: str 需要优化的公式：如'(open - Delay(close, l1)) / Delay(close, l2)'
:param params: dict 需要优化的参数范围：如{"LEN1"：range(1,10,1),"LEN2":range(1,10,1)}
:param name: str (N) 信号的名称
:param price: dataFrame (N) 价格与ret不能同时存在
:param ret: dataFrame (N) 收益
:param high: dataFrame (N) 最高价 用于计算上行收益空间
:param low: dataFrame (N) 最低价 用于计算下行收益空间
:param benchmark_price: dataFrame (N) 基准价格 若不为空收益计算模式为相对benchmark的收益
:param period: int (5) 选股持有期
:param n_quantiles: int (5)
:param mask: 过滤条件 dataFrame (N)
:param can_enter: dataFrame (N) 是否能进场
:param can_exit: dataFrame (N) 是否能出场
:param forward: bool(True) 是否forward return
:param commission: float(0.0008) 手续费率
:param is_event: bool(False) 是否是事件(0/1因子)
:param is_quarterly: bool(False) 是否是季度因子
'''
1. 因子参数优化
price = dv.get_ts('close_adj')high = dv.get_ts('high_adj')low = dv.get_ts('low_adj')price_bench = dv.data_benchmarkoptimizer = Optimizer(dataview=dv,formula='- Correlation(vwap_adj, volume, LEN)',params={"LEN":range(2,15,1)},name='divert',price=price,high=high,low=low,benchmark_price=None,#=None求绝对收益 #=price_bench求相对收益period=30,n_quantiles=5,mask=mask,can_enter=can_enter,can_exit=can_exit,commission=0.0008,#手续费 默认0.0008is_event=False,#是否是事件(0/1因子)is_quarterly=False)#是否是季度因子 默认为False
优化目标
target_type:
- ic类:
return_ic/upside_ret_ic/downside_ret_ic - 持有收益类
long_ret/short_ret/long_short_ret/top_quantile_ret/bottom_quantile_ret/tmb_ret - 收益空间类
long_space/short_space/long_short_space/top_quantile_space/bottom_quantile_space/tmb_space
target:
- ic类
"IC Mean", "IC Std.", "t-stat(IC)", "p-value(IC)", "IC Skew", "IC Kurtosis", "Ann. IR" - 持有收益类
't-stat', "p-value", "skewness", "kurtosis", "Ann. Ret", "Ann. Vol", "Ann. IR", "occurance" - 收益空间类
'Up_sp Mean','Up_sp Std','Up_sp IR','Up_sp Pct5', 'Up_sp Pct25 ','Up_sp Pct50 ', 'Up_sp Pct75','Up_sp Pct95','Up_sp Occur','Down_sp Mean','Down_sp Std', 'Down_sp IR', 'Down_sp Pct5','Down_sp Pct25 ','Down_sp Pct50 ','Down_sp Pct75', 'Down_sp Pct95','Down_sp Occur'
ret_best = optimizer.enumerate_optimizer(target_type="top_quantile_ret",#优化目标类型target="Ann. IR",#优化目标in_sample_range=[20140101,20160101],#样本内范围 默认为None,在全样本上优化ascending=False)#是否按优化目标升序排列(从小到大)
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
# 样本内最优绩效print(ret_best[0]["signal_name"])print(ret_best[0]["ret"])print(ret_best[0]["ic"])print(ret_best[0]["space"])
divert{'LEN': 12}
long_ret short_ret long_short_ret top_quantile_ret \
t-stat 5.110799 -4.669524 3.005345 33.349300
p-value 0.000000 0.000000 0.002790 0.000000
skewness -0.575172 0.309078 -1.131889 0.675351
kurtosis -0.018146 -0.320506 4.012586 3.707588
Ann. Ret 0.262372 -0.254147 0.026183 0.308727
Ann. Vol 0.398883 0.422892 0.067692 0.534983
Ann. IR 0.657766 -0.600973 0.386791 0.577079
occurance 488.000000 488.000000 488.000000 26941.000000
bottom_quantile_ret tmb_ret all_sample_ret
t-stat 23.930533 4.042585 68.882384
p-value 0.000000 0.000060 0.000000
skewness 0.729822 -0.948665 0.951901
kurtosis 2.493816 3.201947 4.295649
Ann. Ret 0.232219 0.078149 0.299248
Ann. Vol 0.564872 0.150204 0.563383
Ann. IR 0.411100 0.520286 0.531163
occurance 27335.000000 488.000000 135662.000000
return_ic upside_ret_ic downside_ret_ic
IC Mean 4.960313e-02 0.007377 6.425467e-02
IC Std. 1.375747e-01 0.136801 1.358379e-01
t-stat(IC) 7.964903e+00 1.191293 1.044946e+01
p-value(IC) 1.177710e-14 0.234119 3.385348e-23
IC Skew -2.779783e-01 -0.167767 2.016567e-01
IC Kurtosis 3.472020e-01 0.873831 3.616037e-01
Ann. IR 3.605542e-01 0.053927 4.730247e-01
long_space short_space long_short_space top_quantile_space \
Up_sp Mean 0.156935 0.113816 0.270751 0.163129
Up_sp Std 0.091596 0.105440 0.113461 0.157858
Up_sp IR 1.713338 1.079438 2.386298 1.033390
Up_sp Pct5 0.044771 0.022594 0.127558 0.008423
Up_sp Pct25 0.087302 0.044775 0.176277 0.054945
Up_sp Pct50 0.135787 0.075946 0.242004 0.118632
Up_sp Pct75 0.216750 0.122701 0.354439 0.222801
Up_sp Pct95 0.346134 0.356583 0.474459 0.468291
Up_sp Occur 488.000000 488.000000 488.000000 26941.000000
Down_sp Mean -0.108474 -0.168088 -0.276562 -0.105026
Down_sp Std 0.103840 0.102438 0.118263 0.116824
Down_sp IR -1.044627 -1.640874 -2.338534 -0.899011
Down_sp Pct5 -0.339163 -0.370044 -0.480706 -0.366672
Down_sp Pct25 -0.109113 -0.233223 -0.369367 -0.133969
Down_sp Pct50 -0.068448 -0.138006 -0.239328 -0.063891
Down_sp Pct75 -0.046176 -0.092070 -0.185622 -0.026318
Down_sp Pct95 -0.022139 -0.043186 -0.136297 -0.003715
Down_sp Occur 488.000000 488.000000 488.000000 26941.000000
bottom_quantile_space tmb_space all_sample_space
Up_sp Mean 0.166828 0.281121 0.167836
Up_sp Std 0.169810 0.110843 0.172231
Up_sp IR 0.982437 2.536211 0.974487
Up_sp Pct5 0.005584 0.135413 0.007070
Up_sp Pct25 0.050753 0.183756 0.052442
Up_sp Pct50 0.115835 0.278703 0.118884
Up_sp Pct75 0.229334 0.349264 0.227149
Up_sp Pct95 0.498775 0.489607 0.492549
Up_sp Occur 27335.000000 488.000000 135662.000000
Down_sp Mean -0.114959 -0.274749 -0.108013
Down_sp Std 0.121342 0.120116 0.117617
Down_sp IR -0.947400 -2.287362 -0.918343
Down_sp Pct5 -0.391534 -0.495471 -0.376228
Down_sp Pct25 -0.149821 -0.359070 -0.139779
Down_sp Pct50 -0.071857 -0.240467 -0.066070
Down_sp Pct75 -0.030352 -0.181938 -0.027373
Down_sp Pct95 -0.004729 -0.129625 -0.004144
Down_sp Occur 27335.000000 488.000000 135662.000000
from jaqs_fxdayu.research import SignalDiggerfrom jaqs_fxdayu.research.signaldigger.analysis import analysisimport matplotlib.pyplot as pltobj = SignalDigger()def draw_analysis(signal_data,period):obj.signal_data = signal_dataobj.period = periodobj.create_full_report()plt.show()
全样本可视化
draw_analysis(optimizer.all_signals[ret_best[0]["signal_name"]], period=30)
Value of signals of Different Quantiles Statistics
min max mean std count count %
quantile
1 -3.062545 -0.054923 -0.790034 0.111886 51786 20.144786
2 -0.910108 0.444112 -0.605170 0.154872 51418 20.001634
3 -0.852226 0.664258 -0.439884 0.198861 51393 19.991909
4 -0.772911 0.801471 -0.239247 0.232510 51418 20.001634
5 -0.644317 0.987630 0.146103 0.293025 51054 19.860038
Figure saved: /home/xinger/Desktop/qtc/多因子选股模型/returns_report.pdf
Information Analysis
ic
IC Mean 0.059
IC Std. 0.133
t-stat(IC) 13.244
p-value(IC) 0.000
IC Skew -0.084
IC Kurtosis 0.237
Ann. IR 0.445
Figure saved: /home/xinger/Desktop/qtc/多因子选股模型/information_report.pdf
<matplotlib.figure.Figure at 0x7f9a7c4107b8>
样本内可视化
draw_analysis(optimizer.all_signals[ret_best[0]["signal_name"]].loc[20140101:20160101], period=30)
Value of signals of Different Quantiles Statistics
min max mean std count count %
quantile
1 -2.080126 -0.054923 -0.802377 0.108587 27335 20.149342
2 -0.910108 0.444112 -0.629030 0.159256 27131 19.998968
3 -0.852226 0.664258 -0.467972 0.211465 27124 19.993808
4 -0.772911 0.801471 -0.269276 0.249928 27131 19.998968
5 -0.644317 0.964493 0.119820 0.308311 26941 19.858914
Figure saved: /home/xinger/Desktop/qtc/多因子选股模型/returns_report.pdf
Information Analysis
ic
IC Mean 0.050
IC Std. 0.138
t-stat(IC) 7.965
p-value(IC) 0.000
IC Skew -0.278
IC Kurtosis 0.347
Ann. IR 0.361
Figure saved: /home/xinger/Desktop/qtc/多因子选股模型/information_report.pdf
<matplotlib.figure.Figure at 0x7f9a7f534dd8>
样本外可视化
draw_analysis(optimizer.all_signals[ret_best[0]["signal_name"]].loc[20160101:], period=30)
Value of signals of Different Quantiles Statistics
min max mean std count count %
quantile
1 -3.062545 -0.240942 -0.776234 0.113892 24451 20.139695
2 -0.906725 0.089250 -0.578515 0.145263 24287 20.004613
3 -0.826314 0.291378 -0.408492 0.178606 24269 19.989786
4 -0.744365 0.490119 -0.205701 0.206260 24287 20.004613
5 -0.616987 0.987630 0.175469 0.271961 24113 19.861293
Figure saved: /home/xinger/Desktop/qtc/多因子选股模型/returns_report.pdf
Information Analysis
ic
IC Mean 0.071
IC Std. 0.126
t-stat(IC) 11.197
p-value(IC) 0.000
IC Skew 0.272
IC Kurtosis -0.239
Ann. IR 0.561
Figure saved: /home/xinger/Desktop/qtc/多因子选股模型/information_report.pdf
<matplotlib.figure.Figure at 0x7f9a7f65b470>
# 以持有期mean_ic为最优化目标ic_best = optimizer.enumerate_optimizer(target_type="return_ic",#优化目标类型target = "IC Mean",in_sample_range=[20140101,20160101],#样本内范围 默认为None,在全样本上优化ascending=False)
print(ic_best[0]["signal_name"])print(ic_best[0]["ic"])print(ic_best[0]["ret"])print(ic_best[0]["space"])
divert{'LEN': 12}
return_ic upside_ret_ic downside_ret_ic
IC Mean 4.960313e-02 0.007377 6.425467e-02
IC Std. 1.375747e-01 0.136801 1.358379e-01
t-stat(IC) 7.964903e+00 1.191293 1.044946e+01
p-value(IC) 1.177710e-14 0.234119 3.385348e-23
IC Skew -2.779783e-01 -0.167767 2.016567e-01
IC Kurtosis 3.472020e-01 0.873831 3.616037e-01
Ann. IR 3.605542e-01 0.053927 4.730247e-01
long_ret short_ret long_short_ret top_quantile_ret \
t-stat 5.110799 -4.669524 3.005345 33.349300
p-value 0.000000 0.000000 0.002790 0.000000
skewness -0.575172 0.309078 -1.131889 0.675351
kurtosis -0.018146 -0.320506 4.012586 3.707588
Ann. Ret 0.262372 -0.254147 0.026183 0.308727
Ann. Vol 0.398883 0.422892 0.067692 0.534983
Ann. IR 0.657766 -0.600973 0.386791 0.577079
occurance 488.000000 488.000000 488.000000 26941.000000
bottom_quantile_ret tmb_ret all_sample_ret
t-stat 23.930533 4.042585 68.882384
p-value 0.000000 0.000060 0.000000
skewness 0.729822 -0.948665 0.951901
kurtosis 2.493816 3.201947 4.295649
Ann. Ret 0.232219 0.078149 0.299248
Ann. Vol 0.564872 0.150204 0.563383
Ann. IR 0.411100 0.520286 0.531163
occurance 27335.000000 488.000000 135662.000000
long_space short_space long_short_space top_quantile_space \
Up_sp Mean 0.156935 0.113816 0.270751 0.163129
Up_sp Std 0.091596 0.105440 0.113461 0.157858
Up_sp IR 1.713338 1.079438 2.386298 1.033390
Up_sp Pct5 0.044771 0.022594 0.127558 0.008423
Up_sp Pct25 0.087302 0.044775 0.176277 0.054945
Up_sp Pct50 0.135787 0.075946 0.242004 0.118632
Up_sp Pct75 0.216750 0.122701 0.354439 0.222801
Up_sp Pct95 0.346134 0.356583 0.474459 0.468291
Up_sp Occur 488.000000 488.000000 488.000000 26941.000000
Down_sp Mean -0.108474 -0.168088 -0.276562 -0.105026
Down_sp Std 0.103840 0.102438 0.118263 0.116824
Down_sp IR -1.044627 -1.640874 -2.338534 -0.899011
Down_sp Pct5 -0.339163 -0.370044 -0.480706 -0.366672
Down_sp Pct25 -0.109113 -0.233223 -0.369367 -0.133969
Down_sp Pct50 -0.068448 -0.138006 -0.239328 -0.063891
Down_sp Pct75 -0.046176 -0.092070 -0.185622 -0.026318
Down_sp Pct95 -0.022139 -0.043186 -0.136297 -0.003715
Down_sp Occur 488.000000 488.000000 488.000000 26941.000000
bottom_quantile_space tmb_space all_sample_space
Up_sp Mean 0.166828 0.281121 0.167836
Up_sp Std 0.169810 0.110843 0.172231
Up_sp IR 0.982437 2.536211 0.974487
Up_sp Pct5 0.005584 0.135413 0.007070
Up_sp Pct25 0.050753 0.183756 0.052442
Up_sp Pct50 0.115835 0.278703 0.118884
Up_sp Pct75 0.229334 0.349264 0.227149
Up_sp Pct95 0.498775 0.489607 0.492549
Up_sp Occur 27335.000000 488.000000 135662.000000
Down_sp Mean -0.114959 -0.274749 -0.108013
Down_sp Std 0.121342 0.120116 0.117617
Down_sp IR -0.947400 -2.287362 -0.918343
Down_sp Pct5 -0.391534 -0.495471 -0.376228
Down_sp Pct25 -0.149821 -0.359070 -0.139779
Down_sp Pct50 -0.071857 -0.240467 -0.066070
Down_sp Pct75 -0.030352 -0.181938 -0.027373
Down_sp Pct95 -0.004729 -0.129625 -0.004144
Down_sp Occur 27335.000000 488.000000 135662.000000
excel_data = optimizer.all_signals[ret_best[0]["signal_name"]][optimizer.all_signals[ret_best[0]["signal_name"]]['quantile']==5]["quantile"].unstack().replace(np.nan, 0).replace(5, 1)print (excel_data.head())excel_data.to_excel('./divert_opt_quantile_5.xlsx')
symbol 000001.SZ 000002.SZ 000008.SZ 000009.SZ 000012.SZ 000024.SZ \
trade_date
20140103 1.0 0.0 0.0 0.0 0.0 1.0
20140106 0.0 0.0 0.0 0.0 0.0 1.0
20140107 0.0 1.0 0.0 0.0 1.0 1.0
20140108 0.0 1.0 0.0 0.0 1.0 1.0
20140109 0.0 0.0 0.0 0.0 0.0 1.0
symbol 000027.SZ 000039.SZ 000046.SZ 000060.SZ ... 601992.SH \
trade_date ...
20140103 0.0 0.0 1.0 0.0 ... 0.0
20140106 0.0 0.0 1.0 0.0 ... 1.0
20140107 0.0 0.0 1.0 0.0 ... 0.0
20140108 0.0 0.0 0.0 0.0 ... 0.0
20140109 0.0 0.0 0.0 0.0 ... 1.0
symbol 601997.SH 601998.SH 603000.SH 603160.SH 603288.SH 603699.SH \
trade_date
20140103 0.0 1.0 0.0 0.0 0.0 0.0
20140106 0.0 0.0 0.0 0.0 0.0 0.0
20140107 0.0 0.0 0.0 0.0 0.0 0.0
20140108 0.0 0.0 0.0 0.0 0.0 0.0
20140109 0.0 0.0 0.0 0.0 0.0 0.0
symbol 603858.SH 603885.SH 603993.SH
trade_date
20140103 0.0 0.0 1.0
20140106 0.0 0.0 0.0
20140107 0.0 0.0 0.0
20140108 0.0 0.0 0.0
20140109 0.0 0.0 0.0
[5 rows x 456 columns]
2. 事件参数优化
event_opt = Optimizer(dataview=dv,formula="(Ts_Mean(close_adj, SHORT)>=Ts_Mean(close_adj, LONG))&&(Delay(Ts_Mean(close_adj, SHORT)<Ts_Mean(close_adj, LONG), 1))",params={'SHORT':range(5,11,1),'LONG':range(30,61,5)},name='cross',price=price,high=high,low=low,benchmark_price=None,#=None求绝对收益 #=price_bench求相对收益period=30,n_quantiles=1,mask=mask,can_enter=can_enter,can_exit=can_exit,commission=0.0008,#手续费 默认0.0008is_event=True,#是否是事件(0/1因子)is_quarterly=False)#是否是季度因子 默认为False
event_best = event_opt.enumerate_optimizer(target_type="long_ret",target="Ann. IR",in_sample_range=[20140101,20160101],#样本内范围 默认为None,在全样本上优化ascending=False)
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 58%
Nan Data Count (should be zero) : 0; Percentage of effective data: 58%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 58%
Nan Data Count (should be zero) : 0; Percentage of effective data: 58%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 58%
Nan Data Count (should be zero) : 0; Percentage of effective data: 58%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 58%
Nan Data Count (should be zero) : 0; Percentage of effective data: 58%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 58%
Nan Data Count (should be zero) : 0; Percentage of effective data: 58%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 58%
Nan Data Count (should be zero) : 0; Percentage of effective data: 58%
# 事件样本内最优绩效 ps：事件没有ic分析结果# 可以进一步尝试优化space，辅以更精细的择时捕捉事件收益print(event_best[0]["signal_name"])print(event_best[0]["ret"])print(event_best[0]["space"])
cross{'SHORT': 10, 'LONG': 55}
long_ret long_short_ret all_sample_ret
t-stat 12.373617 -2.253417 69.704793
p-value 0.000000 0.024880 0.000000
skewness 1.005173 0.410379 0.929952
kurtosis 4.215504 2.859697 4.188103
Ann. Ret 0.456521 -0.047507 0.311164
Ann. Vol 0.496016 0.136064 0.569344
Ann. IR 0.920375 -0.349156 0.546531
occurance 1459.000000 337.000000 131218.000000
long_space all_sample_space
Up_sp Mean 0.165738 0.169461
Up_sp Std 0.165765 0.173734
Up_sp IR 0.999838 0.975401
Up_sp Pct5 0.004526 0.006992
Up_sp Pct25 0.049561 0.052758
Up_sp Pct50 0.118951 0.120412
Up_sp Pct75 0.234501 0.229488
Up_sp Pct95 0.475594 0.496829
Up_sp Occur 1459.000000 131218.000000
Down_sp Mean -0.081863 -0.109031
Down_sp Std 0.090621 0.119049
Down_sp IR -0.903356 -0.915854
Down_sp Pct5 -0.291374 -0.379734
Down_sp Pct25 -0.104825 -0.141476
Down_sp Pct50 -0.053169 -0.066069
Down_sp Pct75 -0.023307 -0.027116
Down_sp Pct95 -0.003920 -0.004029
Down_sp Occur 1459.000000 131218.000000
可视化
# 全样本draw_analysis(event_opt.all_signals[event_best[0]["signal_name"]], period=30)
Value of signals of Different Quantiles Statistics
min max mean std count count %
quantile
1 0.0 1.0 0.012066 0.109179 252621 100.0
Figure saved: /home/xinger/Desktop/qtc/多因子选股模型/returns_report.pdf
Information Analysis
ic
IC Mean -0.006
IC Std. 0.066
t-stat(IC) -2.326
p-value(IC) 0.020
IC Skew -0.083
IC Kurtosis 0.033
Ann. IR -0.091
Figure saved: /home/xinger/Desktop/qtc/多因子选股模型/information_report.pdf
<matplotlib.figure.Figure at 0x7f9a7c54fd30>
# 样本内draw_analysis(event_opt.all_signals[event_best[0]["signal_name"]].loc[20140101:20160101], period=30)
Value of signals of Different Quantiles Statistics
min max mean std count count %
quantile
1 0.0 1.0 0.011119 0.104859 131218 100.0
Figure saved: /home/xinger/Desktop/qtc/多因子选股模型/returns_report.pdf
Information Analysis
ic
IC Mean -0.004
IC Std. 0.068
t-stat(IC) -1.135
p-value(IC) 0.257
IC Skew -0.098
IC Kurtosis 0.562
Ann. IR -0.062
Figure saved: /home/xinger/Desktop/qtc/多因子选股模型/information_report.pdf
<matplotlib.figure.Figure at 0x7f9a7f59e898>
# 样本外draw_analysis(event_opt.all_signals[event_best[0]["signal_name"]].loc[20160101:], period=30)
Value of signals of Different Quantiles Statistics
min max mean std count count %
quantile
1 0.0 1.0 0.013089 0.113655 121403 100.0
Figure saved: /home/xinger/Desktop/qtc/多因子选股模型/returns_report.pdf
Information Analysis
ic
IC Mean -0.008
IC Std. 0.065
t-stat(IC) -2.204
p-value(IC) 0.028
IC Skew -0.073
IC Kurtosis -0.674
Ann. IR -0.125
Figure saved: /home/xinger/Desktop/qtc/多因子选股模型/information_report.pdf
<matplotlib.figure.Figure at 0x7f9a7f3b14a8>
3. 拓展：外部自定义带参算法优化
import pandas as pdimport talib as taslp_opt = Optimizer(dataview=dv,price=price,period=30,n_quantiles=5,mask=mask,can_enter=can_enter,can_exit=can_exit,commission=0.0008,#手续费 默认0.0008is_event=False,#是否是事件(0/1因子)is_quarterly=False)#是否是季度因子 默认为Falsesignals_dict = {}for param in range(2,9,1):slope_df = pd.DataFrame({name: -ta.LINEARREG_SLOPE(value.values, param) for name, value in price.iteritems()}, index=price.index)dv.append_df(slope_df,'SLOPE_PARAM='+str(param))signals_dict['SLOPE_PARAM='+str(param)] = slp_opt.cal_signal(dv.get_ts('SLOPE_PARAM='+str(param)))
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
Nan Data Count (should be zero) : 0; Percentage of effective data: 59%
slp_opt.all_signals = signals_dict
# 以持有期mean_ic为最优化目标slp_best = slp_opt.enumerate_optimizer(target_type="return_ic",#优化目标类型target = "IC Mean",in_sample_range=[20140101,20160101],#样本内范围 默认为None,在全样本上优化ascending=False)
# 样本内绩效print(slp_best[0]["signal_name"])print(slp_best[0]["ic"])print(slp_best[0]["ret"])
SLOPE_PARAM=8
return_ic
IC Mean 4.299888e-02
IC Std. 1.617011e-01
t-stat(IC) 5.831988e+00
p-value(IC) 1.007307e-08
IC Skew 5.618541e-02
IC Kurtosis 2.089043e-01
Ann. IR 2.659157e-01
long_ret short_ret long_short_ret top_quantile_ret \
t-stat 3.819634 -1.719541 4.581976 32.283237
p-value 0.000150 0.086160 0.000010 0.000000
skewness -0.248871 0.321936 0.350289 0.931792
kurtosis -0.149910 -0.288344 1.808498 4.240505
Ann. Ret 0.210085 -0.095689 0.052531 0.312823
Ann. Vol 0.424275 0.428817 0.088438 0.555819
Ann. IR 0.495163 -0.223147 0.593990 0.562815
occurance 481.000000 480.000000 481.000000 26542.000000
bottom_quantile_ret tmb_ret all_sample_ret
t-stat 18.988489 5.454760 69.090547
p-value 0.000000 0.000000 0.000000
skewness 0.834395 0.135840 0.941367
kurtosis 3.770243 4.601257 4.245985
Ann. Ret 0.185419 0.133563 0.303882
Ann. Vol 0.564128 0.188879 0.566108
Ann. IR 0.328682 0.707135 0.536791
occurance 26924.000000 481.000000 133636.000000
