复制链接

克隆策略

以Xgboost AI模型选股为例¶

本文主要从模型训练-模型保存-模型调用三个流程完整讲解如何使用Xgboost模型进行AI策略开发

首先是训练模型¶

In [1]:

# 本代码由可视化策略环境自动生成 2021年12月11日 17:17
# 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。


# 回测引擎：初始化函数，只执行一次
def m19_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.options['hold_days'] = 5

# 回测引擎：每日数据处理函数，每天执行一次
def m19_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']
    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.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
    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 m19_prepare_bigquant_run(context):
    pass


m1 = M.instruments.v2(
    start_date='2010-01-01',
    end_date='2015-01-01',
    market='CN_STOCK_A',
    instrument_list='',
    max_count=0
)

m2 = M.advanced_auto_labeler.v2(
    instruments=m1.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
)

m3 = M.input_features.v1(
    features="""# #号开始的表示注释
# 多个特征，每行一个，可以包含基础特征和衍生特征
return_5
return_10
return_20
avg_amount_0/avg_amount_5
avg_amount_5/avg_amount_20
rank_avg_amount_0/rank_avg_amount_5
rank_avg_amount_5/rank_avg_amount_10
rank_return_0
rank_return_5
rank_return_10
rank_return_0/rank_return_5
rank_return_5/rank_return_10
pe_ttm_0
"""
)

m15 = M.general_feature_extractor.v7(
    instruments=m1.data,
    features=m3.data,
    start_date='',
    end_date='',
    before_start_days=0
)

m16 = M.derived_feature_extractor.v3(
    input_data=m15.data,
    features=m3.data,
    date_col='date',
    instrument_col='instrument',
    drop_na=False,
    remove_extra_columns=False
)

m7 = M.join.v3(
    data1=m2.data,
    data2=m16.data,
    on='date,instrument',
    how='inner',
    sort=False
)

m13 = M.dropnan.v1(
    input_data=m7.data
)

m9 = M.instruments.v2(
    start_date=T.live_run_param('trading_date', '2015-01-01'),
    end_date=T.live_run_param('trading_date', '2017-01-01'),
    market='CN_STOCK_A',
    instrument_list='',
    max_count=0
)

m17 = M.general_feature_extractor.v7(
    instruments=m9.data,
    features=m3.data,
    start_date='',
    end_date='',
    before_start_days=0
)

m18 = M.derived_feature_extractor.v3(
    input_data=m17.data,
    features=m3.data,
    date_col='date',
    instrument_col='instrument',
    drop_na=False,
    remove_extra_columns=False
)

m14 = M.dropnan.v1(
    input_data=m18.data
)

m20 = M.xgboost.v1(
    training_ds=m13.data,
    features=m3.data,
    predict_ds=m14.data,
    num_boost_round=30,
    objective='排序(pairwise)',
    booster='gbtree',
    max_depth=6,
    key_cols='date,instrument',
    group_col='date',
    nthread=1,
    n_gpus=-1,
    other_train_parameters={}
)

m19 = M.trade.v4(
    instruments=m9.data,
    options_data=m20.predictions,
    start_date='',
    end_date='',
    initialize=m19_initialize_bigquant_run,
    handle_data=m19_handle_data_bigquant_run,
    prepare=m19_prepare_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='000300.SHA'
)

[2021-12-11 17:01:11.895148] INFO: moduleinvoker: instruments.v2 开始运行..

[2021-12-11 17:01:11.916463] INFO: moduleinvoker: 命中缓存

[2021-12-11 17:01:11.918145] INFO: moduleinvoker: instruments.v2 运行完成[0.023021s].

[2021-12-11 17:01:11.930924] INFO: moduleinvoker: advanced_auto_labeler.v2 开始运行..

[2021-12-11 17:01:11.944282] INFO: moduleinvoker: 命中缓存

[2021-12-11 17:01:11.949054] INFO: moduleinvoker: advanced_auto_labeler.v2 运行完成[0.018129s].

[2021-12-11 17:01:11.960063] INFO: moduleinvoker: input_features.v1 开始运行..

[2021-12-11 17:01:12.000015] INFO: moduleinvoker: 命中缓存

[2021-12-11 17:01:12.001769] INFO: moduleinvoker: input_features.v1 运行完成[0.041715s].

[2021-12-11 17:01:12.022316] INFO: moduleinvoker: general_feature_extractor.v7 开始运行..

[2021-12-11 17:01:12.032798] INFO: moduleinvoker: 命中缓存

[2021-12-11 17:01:12.034286] INFO: moduleinvoker: general_feature_extractor.v7 运行完成[0.011998s].

[2021-12-11 17:01:12.047849] INFO: moduleinvoker: derived_feature_extractor.v3 开始运行..

[2021-12-11 17:01:12.062040] INFO: moduleinvoker: 命中缓存

[2021-12-11 17:01:12.063939] INFO: moduleinvoker: derived_feature_extractor.v3 运行完成[0.016089s].

[2021-12-11 17:01:12.086899] INFO: moduleinvoker: join.v3 开始运行..

[2021-12-11 17:01:12.101518] INFO: moduleinvoker: 命中缓存

[2021-12-11 17:01:12.104275] INFO: moduleinvoker: join.v3 运行完成[0.017432s].

[2021-12-11 17:01:12.126958] INFO: moduleinvoker: dropnan.v1 开始运行..

[2021-12-11 17:01:12.142355] INFO: moduleinvoker: 命中缓存

[2021-12-11 17:01:12.144708] INFO: moduleinvoker: dropnan.v1 运行完成[0.017756s].

[2021-12-11 17:01:12.150851] INFO: moduleinvoker: instruments.v2 开始运行..

[2021-12-11 17:01:12.162365] INFO: moduleinvoker: 命中缓存

[2021-12-11 17:01:12.164129] INFO: moduleinvoker: instruments.v2 运行完成[0.013286s].

[2021-12-11 17:01:12.176612] INFO: moduleinvoker: general_feature_extractor.v7 开始运行..

[2021-12-11 17:01:15.817860] INFO: 基础特征抽取: 年份 2015, 特征行数=569698

[2021-12-11 17:01:19.491398] INFO: 基础特征抽取: 年份 2016, 特征行数=641546

[2021-12-11 17:01:21.426157] INFO: 基础特征抽取: 年份 2017, 特征行数=0

[2021-12-11 17:01:21.521232] INFO: 基础特征抽取: 总行数: 1211244

[2021-12-11 17:01:21.526342] INFO: moduleinvoker: general_feature_extractor.v7 运行完成[9.349747s].

[2021-12-11 17:01:21.533223] INFO: moduleinvoker: derived_feature_extractor.v3 开始运行..

[2021-12-11 17:01:23.720103] INFO: derived_feature_extractor: 提取完成 avg_amount_0/avg_amount_5, 0.005s

[2021-12-11 17:01:23.725806] INFO: derived_feature_extractor: 提取完成 avg_amount_5/avg_amount_20, 0.004s

[2021-12-11 17:01:23.730033] INFO: derived_feature_extractor: 提取完成 rank_avg_amount_0/rank_avg_amount_5, 0.003s

[2021-12-11 17:01:23.733911] INFO: derived_feature_extractor: 提取完成 rank_avg_amount_5/rank_avg_amount_10, 0.003s

[2021-12-11 17:01:23.737809] INFO: derived_feature_extractor: 提取完成 rank_return_0/rank_return_5, 0.003s

[2021-12-11 17:01:23.742061] INFO: derived_feature_extractor: 提取完成 rank_return_5/rank_return_10, 0.003s

[2021-12-11 17:01:24.840759] INFO: derived_feature_extractor: /y_2015, 569698

[2021-12-11 17:01:26.373523] INFO: derived_feature_extractor: /y_2016, 641546

[2021-12-11 17:01:26.904860] INFO: moduleinvoker: derived_feature_extractor.v3 运行完成[5.37162s].

[2021-12-11 17:01:26.914410] INFO: moduleinvoker: dropnan.v1 开始运行..

[2021-12-11 17:01:27.700984] INFO: dropnan: /y_2015, 565146/569698

[2021-12-11 17:01:28.549920] INFO: dropnan: /y_2016, 636912/641546

[2021-12-11 17:01:28.652515] INFO: dropnan: 行数: 1202058/1211244

[2021-12-11 17:01:28.663127] INFO: moduleinvoker: dropnan.v1 运行完成[1.748712s].

[2021-12-11 17:01:28.684180] INFO: moduleinvoker: xgboost.v1 开始运行..

[2021-12-11 17:10:03.996946] INFO: moduleinvoker: xgboost.v1 运行完成[515.312743s].

[2021-12-11 17:10:06.626183] INFO: moduleinvoker: backtest.v8 开始运行..

[2021-12-11 17:10:06.633161] INFO: backtest: biglearning backtest:V8.6.0

[2021-12-11 17:10:06.635719] INFO: backtest: product_type:stock by specified

[2021-12-11 17:10:06.769895] INFO: moduleinvoker: cached.v2 开始运行..

[2021-12-11 17:10:06.783999] INFO: moduleinvoker: 命中缓存

[2021-12-11 17:10:06.786976] INFO: moduleinvoker: cached.v2 运行完成[0.017111s].

[2021-12-11 17:10:09.006282] INFO: algo: TradingAlgorithm V1.8.6

[2021-12-11 17:10:10.198590] INFO: algo: trading transform...

[2021-12-11 17:10:28.573781] INFO: Performance: Simulated 488 trading days out of 488.

[2021-12-11 17:10:28.575863] INFO: Performance: first open: 2015-01-05 09:30:00+00:00

[2021-12-11 17:10:28.577900] INFO: Performance: last close: 2016-12-30 15:00:00+00:00

[2021-12-11 17:10:35.736990] INFO: moduleinvoker: backtest.v8 运行完成[29.110824s].

[2021-12-11 17:10:35.738558] INFO: moduleinvoker: trade.v4 运行完成[31.718278s].

收益率401.37%
年化收益率129.91%
基准收益率-6.33%
阿尔法1.48
贝塔0.97
夏普比率2.09
胜率0.65
盈亏比0.86
收益波动率42.9%
信息比率0.19
最大回撤49.21%

bigcharts-data-start/{"__type":"tabs","__id":"bigchart-22375e9532854599806ef9d7cf03d760"}/bigcharts-data-end

接下来是模型保存¶

In [7]:

import xgboost as xgb

def prepare_data(df, features, group_col):
    if group_col:
        df.sort_values(group_col, inplace=True)
    try:
        data = df[features].values
        if 'label' in df.columns:
            label = df['label'].values
        else:
            label = None
        dm = xgb.DMatrix(data, label=label)
        if group_col:
            dm.set_group(list(df.groupby(group_col).apply(len)))
        dm.feature_names = features
        return dm
    except KeyError as e:
        log.error('部分特征没有在数据中，执行失败')
        raise e

# 保存模型 
pd.to_pickle(m20.output_model.read(), 'xgboost.csv')
model_info = pd.read_pickle('xgboost.csv') 
features = model_info['features']
parameters = model_info['parameters']
model = model_info['model']
df = m14.data.read()   # 样本外数据
group_col = 'date'
outsample_data = prepare_data(df, features, group_col)

# 预测结果
pred_label = model.predict(outsample_data)
pred_df = pd.DataFrame(
data={'prediction': pred_label}, index=df.index)

# 通过两种方式比较预测结果是否一致
a = m20.predictions.read().sort_values(['date','instrument'])  # 可视化模块的预测结果

df['pred'] = pred_df
b = df[['pred','date','instrument']]  # 源码预测结果

print(a[a['instrument'] =='600000.SHA'].head(50).head(1)) # 随便查看某只股票某天的预测数据是否一致
print(b[b['instrument'] =='600000.SHA'].head(50).head(1))

        prediction       date  instrument
342513    0.323425 2015-01-05  600000.SHA
            pred       date  instrument
342513  0.323425 2015-01-05  600000.SHA

In [6]:

# 查看保存的模型信息
model_info

Out[6]:

{'model': <xgboost.core.Booster at 0x7f76c2be0c40>,
 'features': ['return_5',
  'return_10',
  'return_20',
  'avg_amount_0/avg_amount_5',
  'avg_amount_5/avg_amount_20',
  'rank_avg_amount_0/rank_avg_amount_5',
  'rank_avg_amount_5/rank_avg_amount_10',
  'rank_return_0',
  'rank_return_5',
  'rank_return_10',
  'rank_return_0/rank_return_5',
  'rank_return_5/rank_return_10',
  'pe_ttm_0'],
 'feature_gains':                                  feature  gain
 7                              return_20   280
 6                               return_5   189
 9             avg_amount_5/avg_amount_20   177
 10   rank_avg_amount_0/rank_avg_amount_5   156
 12             avg_amount_0/avg_amount_5   151
 11                              pe_ttm_0   134
 0                          rank_return_5   123
 3                              return_10   122
 8   rank_avg_amount_5/rank_avg_amount_10   119
 2            rank_return_0/rank_return_5   115
 4                         rank_return_10   111
 5           rank_return_5/rank_return_10   104
 1                          rank_return_0    86,
 'parameters': {'objective': 'rank:pairwise',
  'booster': 'gbtree',
  'max_depth': 6,
  'nthread': 1}}