ProStock/src/experiment/regression.py

# %% md
# # LightGBM 回归训练流程（模块化版本）
#
# 使用新的模块化 Trainer 架构，代码更简洁、可维护性更高。
# %% md
# ## 1. 导入依赖
# %%
import os

from src.factors import FactorEngine
from src.training import (
    FactorManager,
    DataPipeline,
    RegressionTask,
    NullFiller,
    Winsorizer,
    StandardScaler,
)
from src.training.core.trainer_v2 import Trainer
from src.training.components.filters import STFilter
from src.experiment.common import (
    SELECTED_FACTORS,
    FACTOR_DEFINITIONS,
    get_label_factor,
    TRAIN_START,
    TRAIN_END,
    VAL_START,
    VAL_END,
    TEST_START,
    TEST_END,
    stock_pool_filter,
    STOCK_FILTER_REQUIRED_COLUMNS,
    OUTPUT_DIR,
    SAVE_PREDICTIONS,
    SAVE_MODEL,
    get_model_save_path,
    save_model_with_factors,
    TOP_N,
)

# 训练类型标识
TRAINING_TYPE = "regression"

# %% md
# ## 2. 训练特定配置
# %%
# Label 配置
LABEL_NAME = "future_return_5"
LABEL_FACTOR = get_label_factor(LABEL_NAME)

# 排除的因子列表
EXCLUDED_FACTORS = [
    "GTJA_alpha010",
    "GTJA_alpha005",
    "GTJA_alpha036",
    "GTJA_alpha027",
    "GTJA_alpha044",
    "GTJA_alpha073",
    "GTJA_alpha104",
    "GTJA_alpha103",
    "GTJA_alpha105",
    "GTJA_alpha092",
    "GTJA_alpha087",
    "GTJA_alpha085",
    "GTJA_alpha062",
    "GTJA_alpha124",
    "GTJA_alpha133",
    "GTJA_alpha131",
    "GTJA_alpha117",
    "GTJA_alpha157",
    "GTJA_alpha162",
    "GTJA_alpha177",
    "GTJA_alpha180",
    "GTJA_alpha191",
]

# 模型参数配置
MODEL_PARAMS = {
    # 基础设置
    "objective": "regression_l1",
    "metric": "mae",
    # 树结构约束
    "max_depth": 5,
    "num_leaves": 24,
    "min_data_in_leaf": 100,
    # 学习参数
    "learning_rate": 0.01,
    "n_estimators": 1500,
    # 随机采样
    "subsample": 0.8,
    "subsample_freq": 1,
    "colsample_bytree": 0.8,
    # 正则化
    "reg_alpha": 0.5,
    "reg_lambda": 1.0,
    # 杂项
    "verbose": -1,
    "random_state": 42,
}

# 日期范围配置
date_range = {
    "train": (TRAIN_START, TRAIN_END),
    "val": (VAL_START, VAL_END),
    "test": (TEST_START, TEST_END),
}

# 输出配置
output_config = {
    "output_dir": OUTPUT_DIR,
    "output_filename": "regression_output.csv",
    "save_predictions": SAVE_PREDICTIONS,
    "save_model": SAVE_MODEL,
    "model_save_path": get_model_save_path(TRAINING_TYPE),
    "top_n": TOP_N,
}


def main():
    """主函数"""
    print("\n" + "=" * 80)
    print("LightGBM 回归模型训练（模块化版本）")
    print("=" * 80)

    # 1. 创建 FactorEngine
    print("\n[1] 创建 FactorEngine")
    engine = FactorEngine()

    # 2. 创建 FactorManager
    print("\n[2] 创建 FactorManager")
    factor_manager = FactorManager(
        selected_factors=SELECTED_FACTORS,
        factor_definitions=FACTOR_DEFINITIONS,
        label_factor=LABEL_FACTOR,
        excluded_factors=EXCLUDED_FACTORS,
    )

    # 3. 创建 DataPipeline
    print("\n[3] 创建 DataPipeline")
    pipeline = DataPipeline(
        factor_manager=factor_manager,
        processor_configs=[
            (NullFiller, {"strategy": "mean"}),
            (Winsorizer, {"lower": 0.01, "upper": 0.99}),
            (StandardScaler, {}),
        ],
        filters=[STFilter(data_router=engine.router)],
        stock_pool_filter_func=stock_pool_filter,
        stock_pool_required_columns=STOCK_FILTER_REQUIRED_COLUMNS,
    )

    # 4. 创建 RegressionTask
    print("\n[4] 创建 RegressionTask")
    task = RegressionTask(
        model_params=MODEL_PARAMS,
        label_name=LABEL_NAME,
    )

    # 5. 创建 Trainer
    print("\n[5] 创建 Trainer")
    trainer = Trainer(
        data_pipeline=pipeline,
        task=task,
        output_config=output_config,
        verbose=True,
    )

    # 6. 执行训练
    print("\n[6] 执行训练")
    results = trainer.run(engine=engine, date_range=date_range)

    # 7. 保存模型和因子信息（如果启用）
    if SAVE_MODEL:
        print("\n[7] 保存模型和因子信息")
        save_model_with_factors(
            model=task.get_model(),
            model_path=output_config["model_save_path"],
            selected_factors=SELECTED_FACTORS,
            factor_definitions=FACTOR_DEFINITIONS,
            fitted_processors=pipeline.get_fitted_processors(),
        )

    print("\n" + "=" * 80)
    print("训练流程完成！")
    print(f"结果保存路径: {os.path.join(OUTPUT_DIR, 'regression_output.csv')}")
    print("=" * 80)

    return results


if __name__ == "__main__":
    main()