feat(training): 实现 LightGBM 模型

- 新增 LightGBMModel:LightGBM 回归模型实现
- 支持自定义参数(objective, num_leaves, learning_rate, n_estimators 等)
- 使用 LightGBM 原生格式保存/加载模型(不依赖 pickle)
- 支持特征重要性提取
- 已注册到 ModelRegistry(@register_model("lightgbm"))
This commit is contained in:
2026-03-03 22:30:37 +08:00
parent 9ca1deae56
commit f35a6a76a6
4 changed files with 432 additions and 0 deletions

View File

@@ -22,6 +22,9 @@ from src.training.components.processors import (
Winsorizer, Winsorizer,
) )
# 模型
from src.training.components.models import LightGBMModel
__all__ = [ __all__ = [
"BaseModel", "BaseModel",
"BaseProcessor", "BaseProcessor",
@@ -31,4 +34,5 @@ __all__ = [
"StandardScaler", "StandardScaler",
"CrossSectionalStandardScaler", "CrossSectionalStandardScaler",
"Winsorizer", "Winsorizer",
"LightGBMModel",
] ]

View File

@@ -0,0 +1,8 @@
"""模型子模块
包含各种机器学习模型的实现。
"""
from src.training.components.models.lightgbm import LightGBMModel
__all__ = ["LightGBMModel"]

View File

@@ -0,0 +1,194 @@
"""LightGBM 模型实现
提供 LightGBM 回归模型的实现,支持特征重要性和原生模型保存。
"""
from typing import Optional
import numpy as np
import pandas as pd
import polars as pl
from src.training.components.base import BaseModel
from src.training.registry import register_model
@register_model("lightgbm")
class LightGBMModel(BaseModel):
"""LightGBM 回归模型
使用 LightGBM 库实现梯度提升回归树。
支持自定义参数、特征重要性提取和原生模型格式保存。
Attributes:
name: 模型名称 "lightgbm"
params: LightGBM 参数字典
model: 训练后的 LightGBM Booster 对象
feature_names_: 特征名称列表
"""
name = "lightgbm"
def __init__(
self,
objective: str = "regression",
metric: str = "rmse",
num_leaves: int = 31,
learning_rate: float = 0.05,
n_estimators: int = 100,
**kwargs,
):
"""初始化 LightGBM 模型
Args:
objective: 目标函数,默认 "regression"
metric: 评估指标,默认 "rmse"
num_leaves: 叶子节点数,默认 31
learning_rate: 学习率,默认 0.05
n_estimators: 迭代次数,默认 100
**kwargs: 其他 LightGBM 参数
"""
self.params = {
"objective": objective,
"metric": metric,
"num_leaves": num_leaves,
"learning_rate": learning_rate,
"verbose": -1, # 抑制训练输出
**kwargs,
}
self.n_estimators = n_estimators
self.model = None
self.feature_names_: Optional[list] = None
def fit(self, X: pl.DataFrame, y: pl.Series) -> "LightGBMModel":
"""训练模型
Args:
X: 特征矩阵 (Polars DataFrame)
y: 目标变量 (Polars Series)
Returns:
self (支持链式调用)
Raises:
ImportError: 未安装 lightgbm
RuntimeError: 训练失败
"""
try:
import lightgbm as lgb
except ImportError:
raise ImportError(
"使用 LightGBMModel 需要安装 lightgbm: pip install lightgbm"
)
# 保存特征名称
self.feature_names_ = X.columns
# 转换为 numpy
X_np = X.to_numpy()
y_np = y.to_numpy()
# 创建数据集
train_data = lgb.Dataset(X_np, label=y_np)
# 训练
self.model = lgb.train(
self.params,
train_data,
num_boost_round=self.n_estimators,
)
return self
def predict(self, X: pl.DataFrame) -> np.ndarray:
"""预测
Args:
X: 特征矩阵 (Polars DataFrame)
Returns:
预测结果 (numpy ndarray)
Raises:
RuntimeError: 模型未训练时调用
"""
if self.model is None:
raise RuntimeError("模型尚未训练,请先调用 fit()")
X_np = X.to_numpy()
return self.model.predict(X_np)
def feature_importance(self) -> Optional[pd.Series]:
"""返回特征重要性
Returns:
特征重要性序列,如果模型未训练则返回 None
"""
if self.model is None or self.feature_names_ is None:
return None
importance = self.model.feature_importance(importance_type="gain")
return pd.Series(importance, index=self.feature_names_)
def save(self, path: str) -> None:
"""保存模型(使用 LightGBM 原生格式)
使用 LightGBM 的原生格式保存,不依赖 pickle
可以在不同环境中加载。
Args:
path: 保存路径
Raises:
RuntimeError: 模型未训练时调用
"""
if self.model is None:
raise RuntimeError("模型尚未训练,无法保存")
self.model.save_model(path)
# 同时保存特征名称LightGBM 原生格式不保存这个)
import json
meta_path = path + ".meta.json"
with open(meta_path, "w") as f:
json.dump(
{
"feature_names": self.feature_names_,
"params": self.params,
"n_estimators": self.n_estimators,
},
f,
)
@classmethod
def load(cls, path: str) -> "LightGBMModel":
"""加载模型
从 LightGBM 原生格式加载模型。
Args:
path: 模型文件路径
Returns:
加载的 LightGBMModel 实例
"""
import lightgbm as lgb
import json
instance = cls()
instance.model = lgb.Booster(model_file=path)
# 加载元数据
meta_path = path + ".meta.json"
try:
with open(meta_path, "r") as f:
meta = json.load(f)
instance.feature_names_ = meta.get("feature_names")
instance.params = meta.get("params", instance.params)
instance.n_estimators = meta.get("n_estimators", instance.n_estimators)
except FileNotFoundError:
# 如果没有元数据文件继续运行feature_names_ 为 None
pass
return instance

View File

@@ -0,0 +1,226 @@
"""测试 LightGBM 模型
验证 LightGBMModel 的训练、预测、保存和加载功能。
"""
import os
import tempfile
import numpy as np
import polars as pl
import pytest
from src.training.components.models.lightgbm import LightGBMModel
class TestLightGBMModel:
"""LightGBMModel 测试类"""
def test_init_default(self):
"""测试默认初始化"""
model = LightGBMModel()
assert model.name == "lightgbm"
assert model.params["objective"] == "regression"
assert model.params["metric"] == "rmse"
assert model.params["num_leaves"] == 31
assert model.params["learning_rate"] == 0.05
assert model.n_estimators == 100
assert model.model is None
def test_init_custom(self):
"""测试自定义参数"""
model = LightGBMModel(
objective="huber",
metric="mae",
num_leaves=50,
learning_rate=0.1,
n_estimators=200,
)
assert model.params["objective"] == "huber"
assert model.params["metric"] == "mae"
assert model.params["num_leaves"] == 50
assert model.params["learning_rate"] == 0.1
assert model.n_estimators == 200
def test_fit_success(self):
"""测试正常训练"""
# 创建简单回归数据
X = pl.DataFrame(
{
"feature1": [1.0, 2.0, 3.0, 4.0, 5.0],
"feature2": [2.0, 4.0, 6.0, 8.0, 10.0],
}
)
y = pl.Series("target", [1.5, 3.0, 4.5, 6.0, 7.5])
model = LightGBMModel(n_estimators=10)
result = model.fit(X, y)
# 验证返回 self支持链式调用
assert result is model
# 验证模型已训练
assert model.model is not None
# 验证特征名称已保存
assert model.feature_names_ == ["feature1", "feature2"]
def test_predict_before_fit(self):
"""测试未训练就预测"""
X = pl.DataFrame(
{
"feature1": [1.0, 2.0],
"feature2": [2.0, 4.0],
}
)
model = LightGBMModel()
with pytest.raises(RuntimeError, match="模型尚未训练"):
model.predict(X)
def test_predict_success(self):
"""测试正常预测"""
# 创建回归数据
np.random.seed(42)
n_samples = 100
X_train = pl.DataFrame(
{
"feature1": np.random.randn(n_samples),
"feature2": np.random.randn(n_samples),
}
)
# y = 2*feature1 + 3*feature2 + noise
y_train = pl.Series(
"target",
2 * X_train["feature1"]
+ 3 * X_train["feature2"]
+ np.random.randn(n_samples) * 0.1,
)
model = LightGBMModel(n_estimators=20, learning_rate=0.1)
model.fit(X_train, y_train)
# 预测新数据(使用明显不同的值)
X_test = pl.DataFrame(
{
"feature1": [-2.0, 3.0],
"feature2": [-1.0, 4.0],
}
)
predictions = model.predict(X_test)
# 验证预测结果格式
assert isinstance(predictions, np.ndarray)
assert len(predictions) == 2
# 验证预测值是数值
assert all(np.isfinite(predictions))
# 验证单调性(第二个样本的 feature 值更大,预测值也应更大)
assert predictions[1] > predictions[0]
def test_feature_importance_before_fit(self):
"""测试未训练就获取特征重要性"""
model = LightGBMModel()
assert model.feature_importance() is None
def test_feature_importance_after_fit(self):
"""测试训练后获取特征重要性"""
X = pl.DataFrame(
{
"feature1": np.random.randn(100),
"feature2": np.random.randn(100),
}
)
y = pl.Series("target", X["feature1"] * 2 + X["feature2"] * 0.1)
model = LightGBMModel(n_estimators=10)
model.fit(X, y)
importance = model.feature_importance()
# 验证特征重要性格式
assert importance is not None
assert len(importance) == 2
assert "feature1" in importance.index
assert "feature2" in importance.index
# feature1 的系数更大,重要性应该更高
assert importance["feature1"] >= importance["feature2"]
def test_save_before_fit(self):
"""测试未训练就保存"""
model = LightGBMModel()
with pytest.raises(RuntimeError, match="模型尚未训练"):
model.save("dummy.txt")
def test_save_and_load(self):
"""测试保存和加载"""
# 训练模型
X = pl.DataFrame(
{
"feature1": [1.0, 2.0, 3.0, 4.0, 5.0],
"feature2": [2.0, 4.0, 6.0, 8.0, 10.0],
}
)
y = pl.Series("target", [2.0, 4.0, 6.0, 8.0, 10.0])
model = LightGBMModel(n_estimators=10, learning_rate=0.1)
model.fit(X, y)
# 保存前预测
X_test = pl.DataFrame(
{
"feature1": [6.0],
"feature2": [12.0],
}
)
pred_before = model.predict(X_test)
# 保存到临时文件
with tempfile.TemporaryDirectory() as tmpdir:
save_path = os.path.join(tmpdir, "model.txt")
model.save(save_path)
# 加载模型
loaded_model = LightGBMModel.load(save_path)
# 验证加载后预测结果相同
pred_after = loaded_model.predict(X_test)
assert pred_after[0] == pytest.approx(pred_before[0], rel=1e-5)
# 验证元数据已恢复
assert loaded_model.feature_names_ == ["feature1", "feature2"]
assert loaded_model.n_estimators == 10
def test_registration(self):
"""测试模型已注册到 registry"""
from src.training.registry import ModelRegistry
model_class = ModelRegistry.get_model("lightgbm")
assert model_class is LightGBMModel
def test_fit_predict_consistency(self):
"""测试多次预测结果一致"""
X = pl.DataFrame(
{
"feature1": np.random.randn(50),
"feature2": np.random.randn(50),
}
)
y = pl.Series("target", X["feature1"] + X["feature2"])
model = LightGBMModel(n_estimators=10)
model.fit(X, y)
X_test = pl.DataFrame(
{
"feature1": [1.0, 2.0, 3.0],
"feature2": [1.0, 2.0, 3.0],
}
)
# 多次预测应该返回相同结果
pred1 = model.predict(X_test)
pred2 = model.predict(X_test)
np.testing.assert_array_almost_equal(pred1, pred2)
if __name__ == "__main__":
pytest.main([__file__, "-v"])