多品种模型

到目前为止，模型训练都是为每个股票代码单独训练一个实例。而现在，也可以在行为相似的品种（例如同一行业的股票）之间训练一个共享模型。

PyBroker v2 支持在传递给 Strategy.add_execution 的所有品种上训练同一个模型，本文档将对此进行演示。

import pybroker
from pybroker import Strategy, YFinance
from sklearn.linear_model import LinearRegression

pybroker.enable_data_source_cache("multi_symbol_models")

<pybroker.cache._L1Cache at 0x7f0d105fee40>

在多个品种上训练一个模型

本文档复用了 训练模型 中的线性回归示例。下面将使用 close_minus_ma 指标，在四只芯片股上训练一个共享的 LinearRegression 模型：

from pybroker.indicator import close_minus_ma

cmma_20 = close_minus_ma("cmma_20", lookback=20, atr_length=14)


def train_slr(symbols, train_data, test_data):
    # Shift within symbols so returns never cross a symbol boundary.
    next_close = train_data.groupby("symbol")["close"].shift(-1)
    train_data["target"] = next_close / train_data["close"] - 1
    train_data = train_data.dropna()
    model = LinearRegression()
    model.fit(train_data[["cmma_20"]], train_data["target"])
    return model, ["cmma_20"]


model_slr = pybroker.model("slr", train_slr, indicators=[cmma_20], pooled=True)

SYMBOLS = ["MU", "TXN", "ADI", "AMAT"]

以 pooled=True 注册模型后，该模型在每次执行中只会训练一次。这会用一个 symbols 元组替换训练函数原本单一的 symbol 参数，并传入这些品种合并后的训练集和测试集划分。

在回测过程中，已训练的模型会在该执行中的所有品种之间共享：

POS_SIZE = 1 / len(SYMBOLS)


def hold_long(ctx):
    pred = ctx.preds("slr")[-1]
    if not ctx.long_pos():
        if pred > 0:
            ctx.buy_shares = ctx.calc_target_shares(POS_SIZE)
    elif pred < 0:
        ctx.sell_all_shares()


strategy = Strategy(YFinance(), start_date="1/1/2021", end_date="1/1/2026")
strategy.add_execution(hold_long, SYMBOLS, models=model_slr)
result = strategy.walkforward(
    warmup=20, windows=3, train_size=0.5, lookahead=1
)
result.metrics_df.head(20)

Backtesting: 2021-01-01 00:00:00 to 2026-01-01 00:00:00

Loading bar data...

[*********************100%***********************]  4 of 4 completed

Loaded bar data: 0:00:00

Computing indicators...

100% (4 of 4) |##########################| Elapsed Time: 0:00:00 Time:  0:00:00

Train split: 2021-01-07 00:00:00 to 2022-04-04 00:00:00
Finished training models: 0:00:00

Test split: 2022-04-05 00:00:00 to 2023-07-05 00:00:00

100% (313 of 313) |######################| Elapsed Time: 0:00:00 Time:  0:00:00

Train split: 2022-04-05 00:00:00 to 2023-07-05 00:00:00
Finished training models: 0:00:00

Test split: 2023-07-06 00:00:00 to 2024-10-01 00:00:00

100% (313 of 313) |######################| Elapsed Time: 0:00:00 Time:  0:00:00

Train split: 2023-07-06 00:00:00 to 2024-10-01 00:00:00
Finished training models: 0:00:00

Test split: 2024-10-02 00:00:00 to 2025-12-31 00:00:00

100% (313 of 313) |######################| Elapsed Time: 0:00:00 Time:  0:00:00

Finished backtest: 0:00:01

	name	value
0	trade_count	170
1	initial_market_value	100000.0
2	end_market_value	173448.08
3	total_pnl	76163.08
4	unrealized_pnl	-2715.0
5	total_return_pct	76.16308
6	total_profit	150256.55
7	total_loss	-74093.47
8	total_fees	0.0
9	max_drawdown	-41802.76
10	max_drawdown_pct	-28.604164
11	max_drawdown_date	2025-04-08 00:00:00
12	win_rate	70.0
13	loss_rate	30.0
14	winning_trades	119
15	losing_trades	51
16	avg_pnl	448.018118
17	avg_return_pct	1.465824
18	avg_trade_bars	14.005882
19	avg_profit	1262.660084