Python | 关于LightGBM

import lightgbm as lgb
from sklearn.metrics import accuracy_score
from sklearn.model_selection import train_test_split
from sklearn.datasets import load_iris
import matplotlib.pyplot as plt

# 初始化
iris = load_iris()
train_X, test_X, train_y, test_y = train_test_split(iris.data, iris.target, test_size=0.2)

# 训练
model = lgb.LGBMClassifier(num_leaves=31, learning_rate=0.05, n_estimators=20)
model.fit(train_X, train_y, eval_set=[(test_X, test_y)], early_stopping_rounds=5)

# 预测与评估
pred_y = model.predict(test_X, num_iteration=model.best_iteration_)
print('Accuracy Score:', accuracy_score(test_y, pred_y))
lgb.plot_importance(model)
plt.show()

import pandas as pd
from sklearn.model_selection import train_test_split
import lightgbm as lgb
from sklearn.metrics import mean_absolute_error, mean_squared_error
from sklearn.impute import SimpleImputer
import numpy as np
import matplotlib.pyplot as plt

# 初始化
df_sample = pd.read_csv('example_data/train.csv')
X = df_sample.drop(['SalePrice'], axis=1).select_dtypes(exclude=['object'])
y = df_sample['SalePrice']
train_X, test_X, train_y, test_y = train_test_split(X.values, y.values, test_size=0.25)

# 一个新知识点：空值处理
imp = SimpleImputer(missing_values=np.nan, strategy='mean')  # 一般填充的strategy都是取均值
train_X = imp.fit_transform(train_X)  # 相当于fit()和transform()连用，先利用train_X对imp进行fit，再处理train_X的空值
test_X = imp.transform(test_X)  # 这里直接transform，即使用train_X的值指导test_X的空值填充

# 训练
model = lgb.LGBMRegressor(objective='regression', num_leaves=31, learning_rate=0.05, n_estimators=20, verbosity=2)
model.fit(train_X, train_y, verbose=False)

# 预测与评估
pred_y = model.predict(test_X)
print('Mean Absolute Error:', mean_absolute_error(pred_y, test_y))  # 平均绝对误差
print('Root Mean Squared Error:', np.sqrt(mean_squared_error(pred_y, test_y)))  # 均方根误差
lgb.plot_importance(model)
plt.show()

import pandas as pd
from sklearn.model_selection import train_test_split, GridSearchCV
import lightgbm as lgb
from sklearn.metrics import mean_absolute_error, mean_squared_error
from sklearn.impute import SimpleImputer
import numpy as np
import matplotlib.pyplot as plt


def grid_optimize(train_X, train_y):
    parameters = {
        'max_depth': [15, 20, 25, 30, 35],
        'learning_rate': [0.01, 0.02, 0.05, 0.1, 0.15],
        'subsample': [0.6, 0.7, 0.8, 0.9, 1],
        'colsample_bytree': [0.6, 0.7, 0.8, 0.9, 1]
    }
    proto_model = lgb.LGBMRegressor(objective='regression')
    gs = GridSearchCV(  # 顺便解释一下GridSearchCV()中的参数
        estimator=proto_model,  # 模型
        param_grid=parameters,  # 待优化参数
        scoring='neg_mean_squared_error',  # 评价标准，默认为None（即使用模型的误差估计方法）
        n_jobs=-1,  # 并行数，默认为1，-1表示使用所有处理器
        cv=3  # 交叉验证参数，默认为5（即五折交叉验证）
    )
    gs.fit(train_X, train_y)
    best_parameters = gs.best_estimator_.get_params()
    for param_name in sorted(parameters.keys()):
        print(f'\n{param_name}: {best_parameters[param_name]}')


def lgb_example(mode='test'):
    # 初始化
    df_sample = pd.read_csv('example_data/train.csv')
    X = df_sample.drop(['SalePrice'], axis=1).select_dtypes(exclude=['object'])
    y = df_sample['SalePrice']
    train_X, test_X, train_y, test_y = train_test_split(X.values, y.values, test_size=0.25)

    # 一个知识点：空值处理
    imp = SimpleImputer(missing_values=np.nan, strategy='mean')  # 一般填充的strategy都是取均值
    train_X = imp.fit_transform(train_X)  # 相当于fit()和transform()连用，先利用train_X对imp进行fit，再处理train_X的空值
    test_X = imp.transform(test_X)  # 这里直接transform，即使用train_X的值指导test_X的空值填充

    if mode == 'test':
        # 参数网格搜索
        grid_optimize(train_X, train_y)
    else:
        # 训练
        model = lgb.LGBMRegressor(  # 这里已经参考了网格搜索的优化结果
            objective='regression',
            num_leaves=31,
            n_estimators=20,
            verbosity=2,
            colsample_bytree=0.7,
            learning_rate=0.05,
            max_depth=25,
            subsample=0.6
        )
        model.fit(train_X, train_y, verbose=False)

        # 预测与评估
        pred_y = model.predict(test_X)
        print('Mean Absolute Error:', mean_absolute_error(pred_y, test_y))  # 平均绝对误差
        print('Root Mean Squared Error:', np.sqrt(mean_squared_error(pred_y, test_y)))  # 均方根误差
        lgb.plot_importance(model)
        plt.show()


if __name__ == '__main__':
    # lgb_example('test')
    lgb_example('operate')