两个结果不一致

h2476 · 2020-12-31 05:59:43 UTC

克隆策略

In [1]:

T.norm([1 / math.log(i + 2) for i in range(0,3)])

Out[1]:

[0.46927872602275644, 0.29608191096586517, 0.23463936301137822]

In [2]:

# 本代码由可视化策略环境自动生成 2020年12月31日 13:17
# 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。


# Python 代码入口函数，input_1/2/3 对应三个输入端，data_1/2/3 对应三个输出端
def m28_run_bigquant_run(input_1, input_2):
    sd = str(input_2.read()['start_date'])
    ed = str(input_2.read()['end_date'])

    dt1 = input_1.read()

    dt1.set_index("date", inplace=True)
    dt1 = dt1[sd:ed]
    dt1 = dt1.reset_index()

    dt1 = DataSource.write_df(dt1)

    return Outputs(data_1=dt1)

# 后处理函数，可选。输入是主函数的输出，可以在这里对数据做处理，或者返回更友好的outputs数据格式。此函数输出不会被缓存。
def m28_post_run_bigquant_run(outputs):
    return outputs

# Python 代码入口函数，input_1/2/3 对应三个输入端，data_1/2/3 对应三个输出端
def m18_run_bigquant_run(input_1, input_2):
    sd = str(input_2.read()['start_date'])
    ed = str(input_2.read()['end_date'])

    dt1 = input_1.read()
    dt1.set_index("date", inplace=True)
    dt1 = dt1[sd:ed]
    dt1 = dt1.reset_index()
    dt1 = DataSource.write_df(dt1)

    return Outputs(data_1=dt1)

# 后处理函数，可选。输入是主函数的输出，可以在这里对数据做处理，或者返回更友好的outputs数据格式。此函数输出不会被缓存。
def m18_post_run_bigquant_run(outputs):
    return outputs

# Python 代码入口函数，input_1/2/3 对应三个输入端，data_1/2/3 对应三个输出端
def m3_run_bigquant_run(input_1, input_2, input_3):
    predictions = input_1.read()

    predictions = predictions.sort_values(['date','score'],ascending = (True,False))

 
    data_1 = DataSource.write_df(predictions)


    return Outputs(data_1=data_1)

# 后处理函数，可选。输入是主函数的输出，可以在这里对数据做处理，或者返回更友好的outputs数据格式。此函数输出不会被缓存。
def m3_post_run_bigquant_run(outputs):
    return outputs

# 回测引擎：初始化函数，只执行一次
def m13_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 =2
    # 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 = 0.6
    context.options['hold_days'] =6
# 回测引擎：每日数据处理函数，每天执行一次
def m13_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.options['hold_days'] # 是否在建仓期间（前 hold_days 天）
    cash_avg = context.portfolio.portfolio_value / context.options['hold_days']  #账户总价值/D
    cash_for_buy = min(context.portfolio.cash, (1 if is_staging else 1.5) * cash_avg)  #账户现金
    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.portfolio.positions.items()}   #仓位

    # 2. 生成卖出订单：hold_days天之后才开始卖出；对持仓的股票，按机器学习算法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.portfolio.positions.items()}
        instruments = list(reversed(list(ranker_prediction.instrument[ranker_prediction.instrument.apply(
                lambda x: x in equities)])))

        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：按机器学习算法预测的排序，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 #账户总价值*c
    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:
            try:
                context.order_value(context.symbol(instrument), cash)
            except:
                return 
# 回测引擎：准备数据，只执行一次
def m13_prepare_bigquant_run(context):
    pass


m19 = M.instruments.v2(
    start_date='2019-07-18',
    end_date='2020-06-19',
    market='CN_STOCK_A',
    instrument_list='',
    max_count=0
)

m20 = 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_5
open
high
close
low
volume
"""
)

m21 = M.instruments.v2(
    start_date='2018-01-01',
    end_date='2020-07-31',
    market='CN_STOCK_A',
    instrument_list='',
    max_count=0
)

m29 = M.instruments.v2(
    start_date='2018-01-01',
    end_date='2019-07-18',
    market='CN_STOCK_A',
    instrument_list='',
    max_count=0
)

m9 = M.advanced_auto_labeler.v2(
    instruments=m29.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/develop/datasource/deprecated/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/develop/bigexpr/usage.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, -3) / shift(open, -1)
shift(close, -7)/close-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
)

m31 = 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_5
open
high
close
low
volume

adjust_factor=1.0

return_0"""
)

m30 = M.use_datasource.v1(
    instruments=m21.data,
    features=m31.data,
    datasource_id='bar1d_CN_STOCK_A',
    start_date='',
    end_date=''
)

m25 = M.derived_feature_extractor.v3(
    input_data=m30.data,
    features=m31.data,
    date_col='date',
    instrument_col='instrument',
    drop_na=True,
    remove_extra_columns=False,
    user_functions={}
)

m33 = M.general_feature_extractor.v7(
    instruments=m21.data,
    features=m31.data,
    start_date='',
    end_date='',
    before_start_days=90
)

m35 = M.derived_feature_extractor.v3(
    input_data=m33.data,
    features=m31.data,
    date_col='date',
    instrument_col='instrument',
    drop_na=True,
    remove_extra_columns=False,
    user_functions={}
)

m34 = M.data_join.v3(
    input_1=m25.data,
    input_2=m35.data,
    on='date,instrument',
    how='inner',
    sort=False
)

m7 = M.dropnan.v2(
    input_data=m34.data
)

m28 = M.cached.v3(
    input_1=m7.data,
    input_2=m29.data,
    run=m28_run_bigquant_run,
    post_run=m28_post_run_bigquant_run,
    input_ports='',
    params='{}',
    output_ports=''
)

m10 = M.data_join.v3(
    input_1=m9.data,
    input_2=m28.data_1,
    on='date,instrument',
    how='inner',
    sort=False
)

m6 = M.stock_ranker_train.v6(
    training_ds=m10.data,
    features=m20.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,
    data_row_fraction=1,
    ndcg_discount_base=1,
    m_lazy_run=False
)

m18 = M.cached.v3(
    input_1=m7.data,
    input_2=m19.data,
    run=m18_run_bigquant_run,
    post_run=m18_post_run_bigquant_run,
    input_ports='',
    params='{}',
    output_ports=''
)

m8 = M.stock_ranker_predict.v5(
    model=m6.model,
    data=m18.data_1,
    m_lazy_run=False
)

m3 = M.cached.v3(
    input_1=m8.predictions,
    run=m3_run_bigquant_run,
    post_run=m3_post_run_bigquant_run,
    input_ports='',
    params='{}',
    output_ports='',
    m_cached=False
)

m13 = M.trade.v4(
    instruments=m19.data,
    options_data=m3.data_1,
    start_date='',
    end_date='',
    initialize=m13_initialize_bigquant_run,
    handle_data=m13_handle_data_bigquant_run,
    prepare=m13_prepare_bigquant_run,
    volume_limit=0,
    order_price_field_buy='open',
    order_price_field_sell='open',
    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='000300.SHA'
)

[2020-12-31 13:11:45.547061] INFO: moduleinvoker: instruments.v2 开始运行..

[2020-12-31 13:11:45.554718] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:11:45.555891] INFO: moduleinvoker: instruments.v2 运行完成[0.008853s].

[2020-12-31 13:11:45.558491] INFO: moduleinvoker: input_features.v1 开始运行..

[2020-12-31 13:11:45.563275] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:11:45.564509] INFO: moduleinvoker: input_features.v1 运行完成[0.006012s].

[2020-12-31 13:11:45.565990] INFO: moduleinvoker: instruments.v2 开始运行..

[2020-12-31 13:11:45.570015] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:11:45.571244] INFO: moduleinvoker: instruments.v2 运行完成[0.005245s].

[2020-12-31 13:11:45.572746] INFO: moduleinvoker: instruments.v2 开始运行..

[2020-12-31 13:11:45.577025] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:11:45.577839] INFO: moduleinvoker: instruments.v2 运行完成[0.005087s].

[2020-12-31 13:11:45.580262] INFO: moduleinvoker: advanced_auto_labeler.v2 开始运行..

[2020-12-31 13:11:45.584309] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:11:45.585632] INFO: moduleinvoker: advanced_auto_labeler.v2 运行完成[0.005365s].

[2020-12-31 13:11:45.587141] INFO: moduleinvoker: input_features.v1 开始运行..

[2020-12-31 13:11:45.591359] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:11:45.592202] INFO: moduleinvoker: input_features.v1 运行完成[0.005063s].

[2020-12-31 13:11:45.599887] INFO: moduleinvoker: use_datasource.v1 开始运行..

[2020-12-31 13:11:45.627610] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:11:45.629282] INFO: moduleinvoker: use_datasource.v1 运行完成[0.029403s].

[2020-12-31 13:11:45.632967] INFO: moduleinvoker: derived_feature_extractor.v3 开始运行..

[2020-12-31 13:11:45.637875] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:11:45.639134] INFO: moduleinvoker: derived_feature_extractor.v3 运行完成[0.006161s].

[2020-12-31 13:11:45.645857] INFO: moduleinvoker: general_feature_extractor.v7 开始运行..

[2020-12-31 13:11:45.650124] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:11:45.651043] INFO: moduleinvoker: general_feature_extractor.v7 运行完成[0.005182s].

[2020-12-31 13:11:45.652612] INFO: moduleinvoker: derived_feature_extractor.v3 开始运行..

[2020-12-31 13:11:45.656581] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:11:45.657415] INFO: moduleinvoker: derived_feature_extractor.v3 运行完成[0.004798s].

[2020-12-31 13:11:45.668048] INFO: moduleinvoker: data_join.v3 开始运行..

[2020-12-31 13:11:45.672779] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:11:45.673987] INFO: moduleinvoker: data_join.v3 运行完成[0.005951s].

[2020-12-31 13:11:45.679635] INFO: moduleinvoker: dropnan.v2 开始运行..

[2020-12-31 13:11:45.683786] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:11:45.684700] INFO: moduleinvoker: dropnan.v2 运行完成[0.005061s].

[2020-12-31 13:11:45.690335] INFO: moduleinvoker: cached.v3 开始运行..

[2020-12-31 13:11:47.488863] INFO: moduleinvoker: cached.v3 运行完成[1.798468s].

[2020-12-31 13:11:47.491875] INFO: moduleinvoker: data_join.v3 开始运行..

[2020-12-31 13:11:50.279724] INFO: moduleinvoker: data_join.v3 运行完成[2.787833s].

[2020-12-31 13:11:50.285894] INFO: moduleinvoker: stock_ranker_train.v6 开始运行..

[2020-12-31 13:11:50.935328] INFO: StockRanker: 特征预处理 ..

[2020-12-31 13:11:51.541753] INFO: StockRanker: prepare data: training ..

[2020-12-31 13:12:01.526524] INFO: StockRanker训练: b04302ba 准备训练: 1258471 行数

[2020-12-31 13:12:01.527859] INFO: StockRanker训练: AI模型训练，将在1258471*6=755.08万数据上对模型训练进行20轮迭代训练。预计将需要3～6分钟。请耐心等待。

[2020-12-31 13:12:01.529106] WARNING: StockRanker训练: 成为高级会员/超级会员，将获得200%～1000%的加速 [url="https://bigquant.com/account/big_member/?from=lab1" style="display: inline-block;padding: 5px 7px;border-radius: 2px;background: #F0BC41;color: white"]快速开通会员[/url]

[2020-12-31 13:12:01.548865] INFO: StockRanker训练: 正在训练 ..

[2020-12-31 13:12:01.618573] INFO: StockRanker训练: 任务状态: Pending

[2020-12-31 13:12:11.666521] INFO: StockRanker训练: 任务状态: Running

[2020-12-31 13:12:21.733808] INFO: StockRanker训练: 00:00:10.9168678, finished iteration 1

[2020-12-31 13:12:31.777817] INFO: StockRanker训练: 00:00:20.5008608, finished iteration 2

[2020-12-31 13:12:41.816942] INFO: StockRanker训练: 00:00:30.6366352, finished iteration 3

[2020-12-31 13:12:51.866974] INFO: StockRanker训练: 00:00:41.6877420, finished iteration 4

[2020-12-31 13:13:01.924452] INFO: StockRanker训练: 00:00:53.5971402, finished iteration 5

[2020-12-31 13:13:11.988834] INFO: StockRanker训练: 00:01:05.5611479, finished iteration 6

[2020-12-31 13:13:32.119082] INFO: StockRanker训练: 00:01:16.9866542, finished iteration 7

[2020-12-31 13:13:42.165181] INFO: StockRanker训练: 00:01:30.8395452, finished iteration 8

[2020-12-31 13:13:52.216712] INFO: StockRanker训练: 00:01:44.2682150, finished iteration 9

[2020-12-31 13:14:12.321296] INFO: StockRanker训练: 00:01:57.6816440, finished iteration 10

[2020-12-31 13:14:22.363574] INFO: StockRanker训练: 00:02:10.0725778, finished iteration 11

[2020-12-31 13:14:32.409337] INFO: StockRanker训练: 00:02:21.9431363, finished iteration 12

[2020-12-31 13:14:42.448487] INFO: StockRanker训练: 00:02:34.3292086, finished iteration 13

[2020-12-31 13:14:52.497275] INFO: StockRanker训练: 00:02:46.0374628, finished iteration 14

[2020-12-31 13:15:12.597687] INFO: StockRanker训练: 00:02:59.1933865, finished iteration 15

[2020-12-31 13:15:22.643268] INFO: StockRanker训练: 00:03:14.0376149, finished iteration 16

[2020-12-31 13:15:42.726648] INFO: StockRanker训练: 00:03:31.0202407, finished iteration 17

[2020-12-31 13:15:52.774613] INFO: StockRanker训练: 00:03:45.6277454, finished iteration 18

[2020-12-31 13:16:12.904470] INFO: StockRanker训练: 00:04:01.0998271, finished iteration 19

[2020-12-31 13:16:22.951199] INFO: StockRanker训练: 00:04:17.0127793, finished iteration 20

[2020-12-31 13:16:32.990594] INFO: StockRanker训练: 任务状态: Succeeded

[2020-12-31 13:16:33.921373] INFO: moduleinvoker: stock_ranker_train.v6 运行完成[283.635463s].

[2020-12-31 13:16:33.924059] INFO: moduleinvoker: cached.v3 开始运行..

[2020-12-31 13:16:35.693356] INFO: moduleinvoker: cached.v3 运行完成[1.769252s].

[2020-12-31 13:16:35.850910] INFO: moduleinvoker: stock_ranker_predict.v5 开始运行..

[2020-12-31 13:16:36.238527] INFO: StockRanker预测: /data ..

[2020-12-31 13:16:38.459327] INFO: moduleinvoker: stock_ranker_predict.v5 运行完成[2.608386s].

[2020-12-31 13:16:38.461567] INFO: moduleinvoker: cached.v3 开始运行..

[2020-12-31 13:16:39.811496] INFO: moduleinvoker: cached.v3 运行完成[1.349884s].

[2020-12-31 13:16:40.867643] INFO: moduleinvoker: backtest.v8 开始运行..

[2020-12-31 13:16:40.872155] INFO: backtest: biglearning backtest:V8.4.2

[2020-12-31 13:16:40.873572] INFO: backtest: product_type:stock by specified

[2020-12-31 13:16:40.975781] INFO: moduleinvoker: cached.v2 开始运行..

[2020-12-31 13:16:40.981223] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:16:40.982289] INFO: moduleinvoker: cached.v2 运行完成[0.006516s].

[2020-12-31 13:16:41.982916] INFO: algo: TradingAlgorithm V1.7.0

[2020-12-31 13:16:42.984361] INFO: algo: trading transform...

[2020-12-31 13:16:45.409653] INFO: Performance: Simulated 225 trading days out of 225.

[2020-12-31 13:16:45.410942] INFO: Performance: first open: 2019-07-18 09:30:00+00:00

[2020-12-31 13:16:45.412371] INFO: Performance: last close: 2020-06-19 15:00:00+00:00

[2020-12-31 13:16:52.079688] INFO: moduleinvoker: backtest.v8 运行完成[11.212059s].

[2020-12-31 13:16:52.080929] INFO: moduleinvoker: trade.v4 运行完成[12.263275s].

设置评估测试数据集，查看训练曲线
[视频教程]StockRanker训练曲线

bigcharts-data-start/{"__type":"tabs","__id":"bigchart-87f232e0840b4804a696be3de3ec4f38"}/bigcharts-data-end

收益率-5.6%
年化收益率-6.25%
基准收益率7.73%
阿尔法-0.12
贝塔0.49
夏普比率-0.54
胜率0.55
盈亏比0.83
收益波动率15.34%
信息比率-0.06
最大回撤14.23%

bigcharts-data-start/{"__type":"tabs","__id":"bigchart-54ab440499c24ab7ac23bb165fd7d416"}/bigcharts-data-end

In [3]:

Out[3]:

In [ ]:

克隆策略

In [14]:

T.norm([1 / math.log(i + 2) for i in range(0,3)])

Out[14]:

[0.46927872602275644, 0.29608191096586517, 0.23463936301137822]

In [15]:

# 本代码由可视化策略环境自动生成 2020年12月31日 13:11
# 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。


# Python 代码入口函数，input_1/2/3 对应三个输入端，data_1/2/3 对应三个输出端
def m28_run_bigquant_run(input_1, input_2):
    sd = str(input_2.read()['start_date'])
    ed = str(input_2.read()['end_date'])

    dt1 = input_1.read()

    dt1.set_index("date", inplace=True)
    dt1 = dt1[sd:ed]
    dt1 = dt1.reset_index()

    dt1 = DataSource.write_df(dt1)

    return Outputs(data_1=dt1)

# 后处理函数，可选。输入是主函数的输出，可以在这里对数据做处理，或者返回更友好的outputs数据格式。此函数输出不会被缓存。
def m28_post_run_bigquant_run(outputs):
    return outputs

# Python 代码入口函数，input_1/2/3 对应三个输入端，data_1/2/3 对应三个输出端
def m18_run_bigquant_run(input_1, input_2):
    sd = str(input_2.read()['start_date'])
    ed = str(input_2.read()['end_date'])

    dt1 = input_1.read()
    dt1.set_index("date", inplace=True)
    dt1 = dt1[sd:ed]
    dt1 = dt1.reset_index()
    dt1 = DataSource.write_df(dt1)

    return Outputs(data_1=dt1)

# 后处理函数，可选。输入是主函数的输出，可以在这里对数据做处理，或者返回更友好的outputs数据格式。此函数输出不会被缓存。
def m18_post_run_bigquant_run(outputs):
    return outputs

# Python 代码入口函数，input_1/2/3 对应三个输入端，data_1/2/3 对应三个输出端
def m3_run_bigquant_run(input_1, input_2, input_3):
    predictions = input_1.read()

    predictions = predictions.sort_values(['date','score'],ascending = (True,False))

 
    data_1 = DataSource.write_df(predictions)


    return Outputs(data_1=data_1)

# 后处理函数，可选。输入是主函数的输出，可以在这里对数据做处理，或者返回更友好的outputs数据格式。此函数输出不会被缓存。
def m3_post_run_bigquant_run(outputs):
    return outputs

# 回测引擎：初始化函数，只执行一次
def m13_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 =2
    # 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 = 0.6
    context.options['hold_days'] =6
# 回测引擎：每日数据处理函数，每天执行一次
def m13_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.options['hold_days'] # 是否在建仓期间（前 hold_days 天）
    cash_avg = context.portfolio.portfolio_value / context.options['hold_days']  #账户总价值/D
    cash_for_buy = min(context.portfolio.cash, (1 if is_staging else 1.5) * cash_avg)  #账户现金
    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.portfolio.positions.items()}   #仓位

    # 2. 生成卖出订单：hold_days天之后才开始卖出；对持仓的股票，按机器学习算法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.portfolio.positions.items()}
        instruments = list(reversed(list(ranker_prediction.instrument[ranker_prediction.instrument.apply(
                lambda x: x in equities)])))

        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：按机器学习算法预测的排序，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 #账户总价值*c
    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:
            try:
                context.order_value(context.symbol(instrument), cash)
            except:
                return 
# 回测引擎：准备数据，只执行一次
def m13_prepare_bigquant_run(context):
    pass


m19 = M.instruments.v2(
    start_date='2019-07-18',
    end_date='2020-06-19',
    market='CN_STOCK_A',
    instrument_list='',
    max_count=0
)

m20 = 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_5
open
high
close
low
volume
"""
)

m21 = M.instruments.v2(
    start_date='2018-01-01',
    end_date='2020-07-31',
    market='CN_STOCK_A',
    instrument_list='',
    max_count=0
)

m29 = M.instruments.v2(
    start_date='2018-01-01',
    end_date='2019-07-18',
    market='CN_STOCK_A',
    instrument_list='',
    max_count=0
)

m9 = M.advanced_auto_labeler.v2(
    instruments=m29.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/develop/datasource/deprecated/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/develop/bigexpr/usage.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, -3) / shift(open, -1)
shift(close, -7)/close-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
)

m31 = 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_5
open
high
close
low
volume

adjust_factor=1.0

return_0"""
)

m30 = M.use_datasource.v1(
    instruments=m21.data,
    features=m31.data,
    datasource_id='bar1d_CN_STOCK_A',
    start_date='',
    end_date=''
)

m25 = M.derived_feature_extractor.v3(
    input_data=m30.data,
    features=m31.data,
    date_col='date',
    instrument_col='instrument',
    drop_na=True,
    remove_extra_columns=False,
    user_functions={}
)

m33 = M.general_feature_extractor.v7(
    instruments=m21.data,
    features=m31.data,
    start_date='',
    end_date='',
    before_start_days=90
)

m35 = M.derived_feature_extractor.v3(
    input_data=m33.data,
    features=m31.data,
    date_col='date',
    instrument_col='instrument',
    drop_na=True,
    remove_extra_columns=False,
    user_functions={}
)

m34 = M.data_join.v3(
    input_1=m25.data,
    input_2=m35.data,
    on='date,instrument',
    how='inner',
    sort=False
)

m7 = M.dropnan.v2(
    input_data=m34.data
)

m28 = M.cached.v3(
    input_1=m7.data,
    input_2=m29.data,
    run=m28_run_bigquant_run,
    post_run=m28_post_run_bigquant_run,
    input_ports='',
    params='{}',
    output_ports=''
)

m10 = M.data_join.v3(
    input_1=m9.data,
    input_2=m28.data_1,
    on='date,instrument',
    how='inner',
    sort=False
)

m6 = M.stock_ranker_train.v6(
    training_ds=m10.data,
    features=m20.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,
    data_row_fraction=1,
    ndcg_discount_base=1,
    m_lazy_run=False
)

m18 = M.cached.v3(
    input_1=m7.data,
    input_2=m19.data,
    run=m18_run_bigquant_run,
    post_run=m18_post_run_bigquant_run,
    input_ports='',
    params='{}',
    output_ports=''
)

m8 = M.stock_ranker_predict.v5(
    model=m6.model,
    data=m18.data_1,
    m_lazy_run=False
)

m3 = M.cached.v3(
    input_1=m8.predictions,
    run=m3_run_bigquant_run,
    post_run=m3_post_run_bigquant_run,
    input_ports='',
    params='{}',
    output_ports='',
    m_cached=False
)

m13 = M.trade.v4(
    options_data=m3.data_1,
    history_ds=m18.data_1,
    start_date='',
    end_date='',
    initialize=m13_initialize_bigquant_run,
    handle_data=m13_handle_data_bigquant_run,
    prepare=m13_prepare_bigquant_run,
    volume_limit=0,
    order_price_field_buy='open',
    order_price_field_sell='open',
    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='000300.SHA'
)

[2020-12-31 13:09:02.880278] INFO: moduleinvoker: instruments.v2 开始运行..

[2020-12-31 13:09:02.884939] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:09:02.885898] INFO: moduleinvoker: instruments.v2 运行完成[0.005621s].

[2020-12-31 13:09:02.887129] INFO: moduleinvoker: input_features.v1 开始运行..

[2020-12-31 13:09:02.891149] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:09:02.892051] INFO: moduleinvoker: input_features.v1 运行完成[0.004915s].

[2020-12-31 13:09:02.893983] INFO: moduleinvoker: instruments.v2 开始运行..

[2020-12-31 13:09:02.899586] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:09:02.900621] INFO: moduleinvoker: instruments.v2 运行完成[0.006632s].

[2020-12-31 13:09:02.902204] INFO: moduleinvoker: instruments.v2 开始运行..

[2020-12-31 13:09:02.906076] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:09:02.907179] INFO: moduleinvoker: instruments.v2 运行完成[0.004965s].

[2020-12-31 13:09:02.909452] INFO: moduleinvoker: advanced_auto_labeler.v2 开始运行..

[2020-12-31 13:09:02.914030] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:09:02.915208] INFO: moduleinvoker: advanced_auto_labeler.v2 运行完成[0.005752s].

[2020-12-31 13:09:02.917140] INFO: moduleinvoker: input_features.v1 开始运行..

[2020-12-31 13:09:02.922067] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:09:02.923009] INFO: moduleinvoker: input_features.v1 运行完成[0.00587s].

[2020-12-31 13:09:02.924474] INFO: moduleinvoker: use_datasource.v1 开始运行..

[2020-12-31 13:09:02.930363] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:09:02.931721] INFO: moduleinvoker: use_datasource.v1 运行完成[0.007237s].

[2020-12-31 13:09:02.933838] INFO: moduleinvoker: derived_feature_extractor.v3 开始运行..

[2020-12-31 13:09:02.993738] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:09:02.994927] INFO: moduleinvoker: derived_feature_extractor.v3 运行完成[0.061085s].

[2020-12-31 13:09:03.000646] INFO: moduleinvoker: general_feature_extractor.v7 开始运行..

[2020-12-31 13:09:03.006866] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:09:03.008013] INFO: moduleinvoker: general_feature_extractor.v7 运行完成[0.007366s].

[2020-12-31 13:09:03.010396] INFO: moduleinvoker: derived_feature_extractor.v3 开始运行..

[2020-12-31 13:09:03.034109] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:09:03.035536] INFO: moduleinvoker: derived_feature_extractor.v3 运行完成[0.025131s].

[2020-12-31 13:09:03.038603] INFO: moduleinvoker: data_join.v3 开始运行..

[2020-12-31 13:09:03.095704] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:09:03.097177] INFO: moduleinvoker: data_join.v3 运行完成[0.058572s].

[2020-12-31 13:09:03.099423] INFO: moduleinvoker: dropnan.v2 开始运行..

[2020-12-31 13:09:03.104670] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:09:03.105822] INFO: moduleinvoker: dropnan.v2 运行完成[0.006401s].

[2020-12-31 13:09:03.112606] INFO: moduleinvoker: cached.v3 开始运行..

[2020-12-31 13:09:03.119057] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:09:03.120238] INFO: moduleinvoker: cached.v3 运行完成[0.007632s].

[2020-12-31 13:09:03.122277] INFO: moduleinvoker: data_join.v3 开始运行..

[2020-12-31 13:09:03.126593] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:09:03.128012] INFO: moduleinvoker: data_join.v3 运行完成[0.005724s].

[2020-12-31 13:09:03.130380] INFO: moduleinvoker: stock_ranker_train.v6 开始运行..

[2020-12-31 13:09:03.327281] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:09:03.371963] INFO: moduleinvoker: stock_ranker_train.v6 运行完成[0.241571s].

[2020-12-31 13:09:03.374721] INFO: moduleinvoker: cached.v3 开始运行..

[2020-12-31 13:09:03.378926] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:09:03.379883] INFO: moduleinvoker: cached.v3 运行完成[0.005165s].

[2020-12-31 13:09:03.382136] INFO: moduleinvoker: stock_ranker_predict.v5 开始运行..

[2020-12-31 13:09:03.391029] INFO: moduleinvoker: 命中缓存

[2020-12-31 13:09:03.391997] INFO: moduleinvoker: stock_ranker_predict.v5 运行完成[0.009864s].

[2020-12-31 13:09:03.394012] INFO: moduleinvoker: cached.v3 开始运行..

[2020-12-31 13:09:04.197484] INFO: moduleinvoker: cached.v3 运行完成[0.803444s].

[2020-12-31 13:09:04.630130] INFO: moduleinvoker: backtest.v8 开始运行..

[2020-12-31 13:09:04.636655] INFO: backtest: biglearning backtest:V8.4.2

[2020-12-31 13:09:04.638040] INFO: backtest: product_type:stock by specified

[2020-12-31 13:09:05.242599] INFO: algo: TradingAlgorithm V1.7.0

[2020-12-31 13:09:05.609837] INFO: algo: trading transform...

[2020-12-31 13:09:08.994374] INFO: Performance: Simulated 225 trading days out of 225.

[2020-12-31 13:09:08.995636] INFO: Performance: first open: 2019-07-18 09:30:00+00:00

[2020-12-31 13:09:08.996841] INFO: Performance: last close: 2020-06-19 15:00:00+00:00

[2020-12-31 13:09:11.652819] INFO: moduleinvoker: backtest.v8 运行完成[7.022697s].

[2020-12-31 13:09:11.654315] INFO: moduleinvoker: trade.v4 运行完成[7.454843s].

设置评估测试数据集，查看训练曲线
[视频教程]StockRanker训练曲线

bigcharts-data-start/{"__type":"tabs","__id":"bigchart-c49448e2f6fa45c7a1d778f2eeded624"}/bigcharts-data-end

收益率20.1%
年化收益率22.77%
基准收益率7.73%
阿尔法0.16
贝塔0.48
夏普比率1.08
胜率0.55
盈亏比1.11
收益波动率17.74%
信息比率0.04
最大回撤10.83%

bigcharts-data-start/{"__type":"tabs","__id":"bigchart-c80992e295df48ca90bfc9a1c1f87af3"}/bigcharts-data-end

In [16]:

Out[16]:

In [ ]:

Daviddddddd · 2020-12-31 08:27:04 UTC

您好，我认真比较了一下两个策略，发现产生不一致的原因是：在自己导入数据做回测的过程中，对于一致涨停、一致跌停的情况没有排除；但是如果直接传入股票代码来使用回测模块，模块会自动过滤掉一致涨跌停的情况而不做交易。具体体现在2019-09-09，可以看到，在手动导入数据的方式中，股票600781.SHA被卖出，而在传入股票代码的回测中，这只股票在这一天没有被卖。实际确认之后可以看到，600781.SHA在2019-09-09是一致跌停的状态。

h2476 · 2020-12-31 08:37:00 UTC

%E6%8D%95%E8%8E%B7

理解了那为什么说回测里采用不同价格的结果是一致的呢?

Daviddddddd · 2020-12-31 09:03:59 UTC

您说的是复权价格吧，那个是说在显示的时候显示的价格，比如真实价格就是你在其他地方都能看到的价格，前复权价格就是以现在的价格为基准反推之前的价格，但是计算收益是不会产生影响的（会做相应调整）。