【案例分享】大盘因子如何构建

qhdxlgd · 2019-06-27 10:10:39 UTC

构建策略时候可能会用到指数的相关因子，因子库中并没有计算。我们可以通过指数特征抽取模块计算大盘因子。

指定指数特征抽取模块中的指数代码，向前抽取天数，通过证券代码列表设置起止时间，通过特征因子列表输入因子表达式，注意指数行情数据只有open，close，high，low，amount和volume基础数据字段供因子表达式构建。

例如构建大盘过去5日收益率： close/shift(close,5)-1，模块会修改因子名称为bm_开头的新因子名称，并输出新的因子列表。

案例：

克隆策略

In [107]:

# 本代码由可视化策略环境自动生成 2019年6月27日 16:33
# 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。


# 回测引擎：初始化函数，只执行一次
def m8_initialize_bigquant_run(context):
    # 加载预测数据
    context.ranker_prediction = context.options['data'].read_df()

    # 系统已经设置了默认的交易手续费和滑点，要修改手续费可使用如下函数
    context.set_commission(PerOrder(buy_cost=0.0003, sell_cost=0.0013, min_cost=5))
    # 预测数据，通过options传入进来，使用 read_df 函数，加载到内存 (DataFrame)
    # 设置买入的股票数量，这里买入预测股票列表排名靠前的5只
    stock_count = 5
    # 每只的股票的权重，如下的权重分配会使得靠前的股票分配多一点的资金，[0.339160, 0.213986, 0.169580, ..]
    context.stock_weights = T.norm([1 / math.log(i + 2) for i in range(0, stock_count)])
    # 设置每只股票占用的最大资金比例
    context.max_cash_per_instrument = 0.2
    context.hold_days = 5

# 回测引擎：每日数据处理函数，每天执行一次
def m8_handle_data_bigquant_run(context, data):
    # 按日期过滤得到今日的预测数据
    ranker_prediction = context.ranker_prediction[
        context.ranker_prediction.date == data.current_dt.strftime('%Y-%m-%d')]

    # 1. 资金分配
    # 平均持仓时间是hold_days，每日都将买入股票，每日预期使用 1/hold_days 的资金
    # 实际操作中，会存在一定的买入误差，所以在前hold_days天，等量使用资金；之后，尽量使用剩余资金（这里设置最多用等量的1.5倍）
    is_staging = context.trading_day_index < context.hold_days # 是否在建仓期间（前 hold_days 天）
    cash_avg = context.portfolio.portfolio_value / context.hold_days
    cash_for_buy = min(context.portfolio.cash, (1 if is_staging else 1.5) * cash_avg)
    cash_for_sell = cash_avg - (context.portfolio.cash - cash_for_buy)
    positions = {e.symbol: p.amount * p.last_sale_price
                 for e, p in context.perf_tracker.position_tracker.positions.items()}

    # 2. 生成卖出订单：hold_days天之后才开始卖出；对持仓的股票，按StockRanker预测的排序末位淘汰
    if not is_staging and cash_for_sell > 0:
        equities = {e.symbol: e for e, p in context.perf_tracker.position_tracker.positions.items()}
        instruments = list(reversed(list(ranker_prediction.instrument[ranker_prediction.instrument.apply(
                lambda x: x in equities and not context.has_unfinished_sell_order(equities[x]))])))
        # print('rank order for sell %s' % instruments)
        for instrument in instruments:
            context.order_target(context.symbol(instrument), 0)
            cash_for_sell -= positions[instrument]
            if cash_for_sell <= 0:
                break

    # 3. 生成买入订单：按StockRanker预测的排序，买入前面的stock_count只股票
    buy_cash_weights = context.stock_weights
    buy_instruments = list(ranker_prediction.instrument[:len(buy_cash_weights)])
    max_cash_per_instrument = context.portfolio.portfolio_value * context.max_cash_per_instrument
    for i, instrument in enumerate(buy_instruments):
        cash = cash_for_buy * buy_cash_weights[i]
        if cash > max_cash_per_instrument - positions.get(instrument, 0):
            # 确保股票持仓量不会超过每次股票最大的占用资金量
            cash = max_cash_per_instrument - positions.get(instrument, 0)
        if cash > 0:
            context.order_value(context.symbol(instrument), cash)

# 回测引擎：准备数据，只执行一次
def m8_prepare_bigquant_run(context):
    pass

# 回测引擎：每个单位时间开始前调用一次，即每日开盘前调用一次。
def m8_before_trading_start_bigquant_run(context, data):
    pass


m5 = 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
return_0
avg_turn_9/2"""
)

m6 = M.instruments.v2(
    start_date='2015-01-01',
    end_date='2018-01-01',
    market='CN_STOCK_A',
    instrument_list='',
    max_count=0
)

m4 = M.general_feature_extractor.v7(
    instruments=m6.data,
    features=m5.data,
    start_date='',
    end_date='',
    before_start_days=90
)

m11 = M.derived_feature_extractor.v3(
    input_data=m4.data,
    features=m5.data,
    date_col='date',
    instrument_col='instrument',
    drop_na=True,
    remove_extra_columns=True,
    user_functions={}
)

m10 = M.advanced_auto_labeler.v2(
    instruments=m6.data,
    label_expr="""# #号开始的表示注释
# 0. 每行一个，顺序执行，从第二个开始，可以使用label字段
# 1. 可用数据字段见 https://bigquant.com/docs/data_history_data.html
#   添加benchmark_前缀，可使用对应的benchmark数据
# 2. 可用操作符和函数见 `表达式引擎 <https://bigquant.com/docs/big_expr.html>`_

# 计算收益：5日收盘价(作为卖出价格)除以明日开盘价(作为买入价格)
shift(close, -5) / shift(open, -1)

# 极值处理：用1%和99%分位的值做clip
clip(label, all_quantile(label, 0.01), all_quantile(label, 0.99))

# 将分数映射到分类，这里使用20个分类
all_wbins(label, 20)

# 过滤掉一字涨停的情况 (设置label为NaN，在后续处理和训练中会忽略NaN的label)
where(shift(high, -1) == shift(low, -1), NaN, label)
""",
    start_date='',
    end_date='',
    benchmark='000300.SHA',
    drop_na_label=True,
    cast_label_int=True,
    user_functions={}
)

m12 = M.join.v3(
    data1=m10.data,
    data2=m11.data,
    on='date,instrument',
    how='inner',
    sort=False
)

m15 = 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
close/shift(close,5)-1
amount+1
ta_sma(close,5)
"""
)

m3 = M.index_feature_extract.v2(
    input_1=m6.data,
    input_2=m15.data,
    before_days=100,
    index='000300.HIX'
)

m1 = M.join.v3(
    data1=m3.data_1,
    data2=m12.data,
    on='date',
    how='inner',
    sort=False
)

m7 = M.features_add.v1(
    input_1=m3.data_2,
    input_2=m5.data
)

m16 = M.instruments.v2(
    start_date='2019-01-01',
    end_date='2019-06-01',
    market='CN_STOCK_A',
    instrument_list='',
    max_count=0
)

m14 = M.index_feature_extract.v2(
    input_1=m16.data,
    input_2=m15.data,
    before_days=100,
    index='000300.HIX'
)

m17 = M.general_feature_extractor.v7(
    instruments=m16.data,
    features=m5.data,
    start_date='',
    end_date='',
    before_start_days=90
)

m18 = M.derived_feature_extractor.v3(
    input_data=m17.data,
    features=m5.data,
    date_col='date',
    instrument_col='instrument',
    drop_na=True,
    remove_extra_columns=True,
    user_functions={}
)

m2 = M.join.v3(
    data1=m18.data,
    data2=m14.data_1,
    on='date',
    how='inner',
    sort=False
)

m9 = M.stock_ranker.v2(
    training_ds=m1.data,
    features=m7.data_1,
    predict_ds=m2.data,
    learning_algorithm='排序',
    number_of_leaves=30,
    minimum_docs_per_leaf=1000,
    number_of_trees=20,
    learning_rate=0.1,
    max_bins=1023,
    feature_fraction=1,
    slim_data=True
)

m8 = M.trade.v4(
    instruments=m16.data,
    options_data=m9.predictions,
    start_date='',
    end_date='',
    initialize=m8_initialize_bigquant_run,
    handle_data=m8_handle_data_bigquant_run,
    prepare=m8_prepare_bigquant_run,
    before_trading_start=m8_before_trading_start_bigquant_run,
    volume_limit=0.025,
    order_price_field_buy='open',
    order_price_field_sell='close',
    capital_base=1000000,
    auto_cancel_non_tradable_orders=True,
    data_frequency='daily',
    price_type='后复权',
    product_type='股票',
    plot_charts=True,
    backtest_only=False,
    benchmark=''
)

2019-06-27 16:31:19.980170 INFO: bigquant: input_features.v1 开始运行.

2019-06-27 16:31:20.279296 INFO: bigquant: 命中缓

2019-06-27 16:31:20.281567 INFO: bigquant: input_features.v1 运行完成[0.3014s]

2019-06-27 16:31:20.285033 INFO: bigquant: instruments.v2 开始运行.

2019-06-27 16:31:20.390050 INFO: bigquant: 命中缓

2019-06-27 16:31:20.392448 INFO: bigquant: instruments.v2 运行完成[0.107398s]

2019-06-27 16:31:20.458586 INFO: bigquant: general_feature_extractor.v7 开始运行.

2019-06-27 16:31:20.555506 INFO: bigquant: 命中缓

2019-06-27 16:31:20.557901 INFO: bigquant: general_feature_extractor.v7 运行完成[0.099326s]

2019-06-27 16:31:20.561928 INFO: bigquant: derived_feature_extractor.v3 开始运行.

2019-06-27 16:31:20.730321 INFO: bigquant: 命中缓

2019-06-27 16:31:20.732696 INFO: bigquant: derived_feature_extractor.v3 运行完成[0.170752s]

2019-06-27 16:31:20.735697 INFO: bigquant: advanced_auto_labeler.v2 开始运行.

2019-06-27 16:31:20.798574 INFO: bigquant: 命中缓

2019-06-27 16:31:20.807942 INFO: bigquant: advanced_auto_labeler.v2 运行完成[0.072238s]

2019-06-27 16:31:20.811297 INFO: bigquant: join.v3 开始运行.

2019-06-27 16:31:20.891029 INFO: bigquant: 命中缓

2019-06-27 16:31:20.896821 INFO: bigquant: join.v3 运行完成[0.085511s]

2019-06-27 16:31:20.900394 INFO: bigquant: input_features.v1 开始运行.

2019-06-27 16:31:20.963844 INFO: bigquant: 命中缓

2019-06-27 16:31:20.966598 INFO: bigquant: input_features.v1 运行完成[0.066193s]

2019-06-27 16:31:20.979865 INFO: bigquant: index_feature_extract.v2 开始运行.

2019-06-27 16:31:21.240578 INFO: bigquant: 命中缓

2019-06-27 16:31:21.242836 INFO: bigquant: index_feature_extract.v2 运行完成[0.262969s]

2019-06-27 16:31:21.246582 INFO: bigquant: join.v3 开始运行.

2019-06-27 16:31:21.394964 INFO: bigquant: 命中缓

2019-06-27 16:31:21.397001 INFO: bigquant: join.v3 运行完成[0.150416s]

2019-06-27 16:31:21.402879 INFO: bigquant: features_add.v1 开始运行.

2019-06-27 16:31:21.536595 INFO: bigquant: 命中缓

2019-06-27 16:31:21.539042 INFO: bigquant: features_add.v1 运行完成[0.13615s]

2019-06-27 16:31:21.542973 INFO: bigquant: instruments.v2 开始运行.

2019-06-27 16:31:21.625088 INFO: bigquant: 命中缓

2019-06-27 16:31:21.627739 INFO: bigquant: instruments.v2 运行完成[0.084748s]

2019-06-27 16:31:21.632878 INFO: bigquant: index_feature_extract.v2 开始运行.

2019-06-27 16:31:22.049900 INFO: bigquant: 命中缓

2019-06-27 16:31:22.051927 INFO: bigquant: index_feature_extract.v2 运行完成[0.41905s]

2019-06-27 16:31:22.154042 INFO: bigquant: general_feature_extractor.v7 开始运行.

2019-06-27 16:31:22.291052 INFO: bigquant: 命中缓

2019-06-27 16:31:22.297695 INFO: bigquant: general_feature_extractor.v7 运行完成[0.143638s]

2019-06-27 16:31:22.300908 INFO: bigquant: derived_feature_extractor.v3 开始运行.

2019-06-27 16:31:22.378766 INFO: bigquant: 命中缓

2019-06-27 16:31:22.381183 INFO: bigquant: derived_feature_extractor.v3 运行完成[0.080265s]

2019-06-27 16:31:22.384514 INFO: bigquant: join.v3 开始运行.

2019-06-27 16:31:22.459934 INFO: bigquant: 命中缓

2019-06-27 16:31:22.465530 INFO: bigquant: join.v3 运行完成[0.080999s]

2019-06-27 16:31:22.472007 INFO: bigquant: stock_ranker.v2 开始运行.

2019-06-27 16:31:22.711792 INFO: bigquant: 命中缓

2019-06-27 16:31:22.981622 INFO: bigquant: stock_ranker.v2 运行完成[0.509601s]

2019-06-27 16:31:23.138535 INFO: bigquant: backtest.v8 开始运行.

2019-06-27 16:31:23.142211 INFO: bigquant: biglearning backtest:V8.2.

2019-06-27 16:31:23.144309 INFO: bigquant: product_type:stock by specifie

2019-06-27 16:31:23.379068 INFO: bigquant: cached.v2 开始运行.

2019-06-27 16:31:23.526450 INFO: bigquant: 命中缓

2019-06-27 16:31:23.528956 INFO: bigquant: cached.v2 运行完成[0.14989s]

2019-06-27 16:31:30.941657 INFO: algo: TradingAlgorithm V1.5.

2019-06-27 16:31:33.787299 INFO: algo: trading transform..

2019-06-27 16:31:37.305038 INFO: Performance: Simulated 99 trading days out of 99

2019-06-27 16:31:37.307593 INFO: Performance: first open: 2019-01-02 09:30:00+00:0

2019-06-27 16:31:37.309895 INFO: Performance: last close: 2019-05-31 15:00:00+00:0

2019-06-27 16:31:46.540824 INFO: bigquant: backtest.v8 运行完成[23.402275s]

StockRanker
特征重要性
模型
NDCG

设置测试数据集，查看训练迭代过程的NDCG

bigcharts-data-start/{"__type":"tabs","__id":"bigchart-c69dca47efb04a6fa3214eed0442d40d"}/bigcharts-data-end

收益率38.34%
年化收益率128.43%
基准收益率20.56%
阿尔法0.51
贝塔0.68
夏普比率3.13
胜率0.58
盈亏比1.55
收益波动率26.61%
信息比率0.1
最大回撤8.48%

bigcharts-data-start/{"__type":"tabs","__id":"bigchart-1a30c20a8d66402a9a008b5f19d6f5be"}/bigcharts-data-end

chenys101 · 2019-06-27 16:36:55 UTC

走向跟大盘好像

chenys101 · 2019-06-30 15:12:20 UTC

试了一下，单边上涨时加入这个大盘因子很有效，17-18年就有相反效果，不知道有没有改进的地方

qhdxlgd · 2019-07-01 02:47:56 UTC

有可能加点别的大盘类的类型因子比如反转之类的，或者别的周期/防御指数的因子或许能有改进，您可以试试，我这里就是给了一个构建大盘因子的例子思路。

wicked_code · 2020-05-04 12:55:13 UTC

现在没用了，取不出数据。