Lightgbm的使用(未完待续)

本文用于记录Lightgbm的使用，使用包含两种，原生lightgbm与以scikit learn的接口方式，下文主要用例子进行说明，具体的api需要看lightgbm的文档，进一步了解还要看论文，理解参数设置的含义。

1.Scikit learn 接口方式

多分类例子

from sklearn.model_selection import train_test_split
from lightgbm.sklearn import LGBMClassifier
from sklearn.datasets import load_iris
from sklearn.metrics import accuracy_score
from sklearn.preprocessing import StandardScaler


iris = load_iris()
data = iris.data
target = iris.target

print(target)

clf = LGBMClassifier(
    num_leaves=30,
    learning_rate=0.05,
    max_depth=5,
    n_estimators=1000,
    objective='multiclass',
    subsample=1,
    # reg_lambda=0.01,
    n_jobs=-1
)

train_x, test_x,train_y,test_y = train_test_split(data,target)
stand = StandardScaler().fit(train_x)
stand_train_x = stand.fit_transform(train_x)
stand_test_x = stand.fit_transform(test_x)

clf.fit(
    X=stand_train_x,
    y=train_y,
    eval_set=(stand_test_x,test_y),
    early_stopping_rounds=200,
    verbose=200
)

result = clf.predict(stand_test_x)
acc = accuracy_score(result, test_y)
print("准确率:{}".format(acc))

回归的例子

from sklearn.model_selection import train_test_split
from lightgbm.sklearn import LGBMClassifier,LGBMRegressor
from sklearn.datasets import load_iris
from sklearn.metrics import mean_squared_error
from sklearn.metrics import accuracy_score
from sklearn.preprocessing import StandardScaler


iris = load_iris()
data = iris.data
target = iris.target

print(target)

clf = LGBMRegressor(
    num_leaves=30,
    learning_rate=0.05,
    max_depth=5,
    n_estimators=1000,
    subsample=1,
    # reg_lambda=0.01,
    n_jobs=-1
)

train_x, test_x,train_y,test_y = train_test_split(data,target)
stand = StandardScaler().fit(train_x)
stand_train_x = stand.fit_transform(train_x)
stand_test_x = stand.fit_transform(test_x)

clf.fit(
    X=stand_train_x,
    y=train_y,
    eval_set=(stand_test_x,test_y),
    eval_metric='l2',
    early_stopping_rounds=200,
    verbose=200
)

result = clf.predict(stand_test_x)
loss = mean_squared_error(result, test_y)
print("损失值:{}".format(loss))