复制链接

克隆策略

案例¶

利用表名__列名实现跨表因子的混合抽取
分别实现滚动2日成交额均值、明日收盘价、(收盘价+开盘价)、每日pb_lf和每日pe_ttm的跨表因子数据提取、合并和表达式计算

In [49]:

# 本代码由可视化策略环境自动生成 2022年11月9日 16:55
# 本代码单元只能在可视化模式下编辑。您也可以拷贝代码，粘贴到新建的代码单元或者策略，然后修改。


m2 = M.instruments.v2(
    start_date='2022-09-10',
    end_date='2022-11-08',
    market='CN_STOCK_A',
    instrument_list="""003003.SZA
#600519.SHA
#003041.SZA

""",
    max_count=0
)

m3 = M.input_features.v1(
    features="""
# #号开始的表示注释，注释需单独一行
# 多个特征，每行一个，可以包含基础特征和衍生特征，特征须为本平台特征
# amount=bar1d_CN_STOCK_A__amount
#mean(amount, 2)
#close_1=shift(bar1d_CN_STOCK_A__close, 1)
#close_avg=(bar1d_CN_STOCK_A__close + bar1d_CN_STOCK_A__open)/2
# pb_lf=market_value_CN_STOCK_A__pb_lf
#pe_ttm=market_value_CN_STOCK_A__pe_ttm
# amount_5
# rank_amount_5
# #((close_0-open_0)/open_0)/((close_0-open_4)/open_4)
# daily_return_3
# # rank_avg_amount_5
# avg_amount_5
# # rank_amount_5
# rank_amount_0
# std(turn_0,10)
# avg_turn_5
# rank_swing_volatility_60_0
# fe1=sum(max(0,high_0-delay((high_0+low_0+close_0)/3,1)),26)/sum(max(0,delay((high_0+low_0+close_0)/3,1)-1),26)*100
# mean(turn_0*(return_0-1),30)
# std(return_0-1,60)
# std(return_0-1,15)
# fe2o=rank(((close_0-sum(min(low_0,delay(close_0,1)),6))/sum(max(high_0,delay(close_0,1))-min(low_0,delay(close_0,1)),6)*12*24+(close_0-sum(min(low_0,delay(close_0,1)),12))/sum(max(high_0,delay(close_0,1))-min(low_0,delay(close_0,1)),12)*6*24+(close_0-sum(min(low_0,delay(close_0,1)),24))/sum(max(high_0,delay(close_0,1))-min(low_0,delay(close_0,1)),24)*6*24)/-20000)
# # rank_amount_0/rank_avg_amount_5
fs_roe_0
fe3=sum((close_0-open_0)/open_0>0.04,20)
rank_volatility_10_0
rank_swing_volatility_10_0
# market_cap_float_0
#fe2=((close_0-sum(min(low_0,delay(close_0,1)),6))/sum(max(high_0,delay(close_0,1))-min(low_0,delay(close_0,1)),6)*12*24+(close_0-sum(min(low_0,delay(close_0,1)),12))/sum(max(high_0,delay(close_0,1))-min(low_0,delay(close_0,1)),12)*6*24+(close_0-sum(min(low_0,delay(close_0,1)),24))/sum(max(high_0,delay(close_0,1))-min(low_0,delay(close_0,1)),24)*6*24)*100/(6*12+6*24+12*24)"""
)

m1 = M.general_feature_extractor.v7(
    instruments=m2.data,
    features=m3.data,
    start_date='2022-05-10',
    end_date='2022-11-08',
    before_start_days=200,
    m_cached=False
)

m4 = M.derived_feature_extractor.v3(
    input_data=m1.data,
    features=m3.data,
    date_col='date',
    instrument_col='instrument',
    drop_na=True,
    remove_extra_columns=True,
    user_functions={}
)

[2022-11-09 16:54:32.856479] INFO: moduleinvoker: instruments.v2 开始运行..

[2022-11-09 16:54:32.867381] INFO: moduleinvoker: 命中缓存

[2022-11-09 16:54:32.870382] INFO: moduleinvoker: instruments.v2 运行完成[0.013892s].

[2022-11-09 16:54:32.876873] INFO: moduleinvoker: input_features.v1 开始运行..

[2022-11-09 16:54:32.886629] INFO: moduleinvoker: 命中缓存

[2022-11-09 16:54:32.889588] INFO: moduleinvoker: input_features.v1 运行完成[0.012712s].

[2022-11-09 16:54:32.927767] INFO: moduleinvoker: general_feature_extractor.v7 开始运行..

[2022-11-09 16:54:33.543849] INFO: 基础特征抽取: 年份 2021, 特征行数=51

[2022-11-09 16:54:34.062153] INFO: 基础特征抽取: 年份 2022, 特征行数=204

[2022-11-09 16:54:34.139004] INFO: 基础特征抽取: 总行数: 255

[2022-11-09 16:54:34.141395] INFO: moduleinvoker: general_feature_extractor.v7 运行完成[1.213615s].

[2022-11-09 16:54:34.203003] INFO: moduleinvoker: derived_feature_extractor.v3 开始运行..

[2022-11-09 16:54:34.284469] INFO: derived_feature_extractor: 提取完成 fe3=sum((close_0-open_0)/open_0>0.04,20), 0.005s

[2022-11-09 16:54:34.384947] INFO: derived_feature_extractor: /y_2021, 51

[2022-11-09 16:54:34.426624] INFO: derived_feature_extractor: /y_2022, 204

[2022-11-09 16:54:34.539269] INFO: moduleinvoker: derived_feature_extractor.v3 运行完成[0.336252s].

In [50]:

#m5.get_attr

In [51]:

m4.data.read()

Out[51]:

	date	instrument	fs_roe_0	rank_volatility_10_0	rank_swing_volatility_10_0	fe3