LightGBM用法速查表
内置方式建模
作者：小菜鸡一号
1.要把数据读取成Dataset格式
2.lgb.train去训练
# coding: utf-8
import json
import lightgbm as lgb
import pandas as pd
from sklearn.metrics import mean_squared_error
# 加载数据集合
print('加载数据...')
df_train = pd.read_csv('./data/regression.train.txt', header=None, sep='\t')
df_test = pd.read_csv('./data/regression.test.txt', header=None, sep='\t')
# 设定训练集和测试集
y_train = df_train[0].values
y_test = df_test[0].values
X_train = df_train.drop(0, axis=1).values
X_test = df_test.drop(0, axis=1).values
# 构建lgb中的Dataset格式
lgb_train = lgb.Dataset(X_train, y_train)
lgb_eval = lgb.Dataset(X_test, y_test, reference=lgb_train)
# 敲定好一组参数
params = {
'task': 'train',
'boosting_type': 'gbdt',
'objective': 'regression',
'metric': {'l2', 'auc'},
'num_leaves': 31,
'learning_rate': 0.05,
'feature_fraction': 0.9,
'bagging_fraction': 0.8,
'bagging_freq': 5,
'verbose': 0
}
print('开始训练...')
# 训练
gbm = lgb.train(params,
lgb_train,
num_boost_round=20,
valid_sets=lgb_eval,
early_stopping_rounds=5)
# 保存模型
print('保存模型...')
# 保存模型到文件中
gbm.save_model('model.txt')
print('开始预测...')
# 预测
y_pred = gbm.predict(X_test, num_iteration=gbm.best_iteration)
# 评估
print('预估结果的rmse为:')
print(mean_squared_error(y_test, y_pred) ** 0.5)
添加样本权重训练
# coding: utf-8
import json
import lightgbm as lgb
import pandas as pd
import numpy as np
from sklearn.metrics import mean_squared_error
import warnings
warnings.filterwarnings("ignore")
# 加载数据集
print('加载数据...')
df_train = pd.read_csv('./data/binary.train', header=None, sep='\t')
df_test = pd.read_csv('./data/binary.test', header=None, sep='\t')
W_train = pd.read_csv('./data/binary.train.weight', header=None)[0]
W_test = pd.read_csv('./data/binary.test.weight', header=None)[0]
y_train = df_train[0].values
y_test = df_test[0].values
X_train = df_train.drop(0, axis=1).values
X_test = df_test.drop(0, axis=1).values
num_train, num_feature = X_train.shape
# 加载数据的同时加载权重
lgb_train = lgb.Dataset(X_train, y_train,
weight=W_train, free_raw_data=False)
lgb_eval = lgb.Dataset(X_test, y_test, reference=lgb_train,
weight=W_test, free_raw_data=False)
# 设定参数
params = {
'boosting_type': 'gbdt',
'objective': 'binary',
'metric': 'binary_logloss',
'num_leaves': 31,
'learning_rate': 0.05,
'feature_fraction': 0.9,
'bagging_fraction': 0.8,
'bagging_freq': 5,
'verbose': 0
}
# 产出特征名称
feature_name = ['feature_' + str(col) for col in range(num_feature)]
print('开始训练...')
gbm = lgb.train(params,
lgb_train,
num_boost_round=10,
valid_sets=lgb_train, # 评估训练集
feature_name=feature_name,
categorical_feature=[21])
模型的载入与预测
# 查看特征名称
print('完成10轮训练...')
print('第7个特征为:')
print(repr(lgb_train.feature_name[6]))
# 存储模型
gbm.save_model('./model/lgb_model.txt')
# 特征名称
print('特征名称:')
print(gbm.feature_name())
# 特征重要度
print('特征重要度:')
print(list(gbm.feature_importance()))
# 加载模型
print('加载模型用于预测')
bst = lgb.Booster(model_file='./model/lgb_model.txt')
# 预测
y_pred = bst.predict(X_test)
# 在测试集评估效果
print('在测试集上的rmse为:')
print(mean_squared_error(y_test, y_pred) ** 0.5)
接着之前的模型继续训练
# 继续训练
# 从./model/model.txt中加载模型初始化
gbm = lgb.train(params,
lgb_train,
num_boost_round=10,
init_model='./model/lgb_model.txt',
valid_sets=lgb_eval)
print('以旧模型为初始化，完成第 10-20 轮训练...')
# 在训练的过程中调整超参数
# 比如这里调整的是学习率
gbm = lgb.train(params,
lgb_train,
num_boost_round=10,
init_model=gbm,
learning_rates=lambda iter: 0.05 * (0.99 ** iter),
valid_sets=lgb_eval)
print('逐步调整学习率完成第 20-30 轮训练...')
# 调整其他超参数
gbm = lgb.train(params,
lgb_train,
num_boost_round=10,
init_model=gbm,
valid_sets=lgb_eval,
callbacks=[lgb.reset_parameter(bagging_fraction=[0.7] * 5 + [0.6] * 5)])
print('逐步调整bagging比率完成第 30-40 轮训练...')
自定义损失函数
# 类似在xgboost中的形式
# 自定义损失函数需要
def loglikelood(preds, train_data):
labels = train_data.get_label()
preds = 1. / (1. + np.exp(-preds))
grad = preds - labels
hess = preds * (1. - preds)
return grad, hess
# 自定义评估函数
def binary_error(preds, train_data):
labels = train_data.get_label()
return 'error', np.mean(labels != (preds > 0.5)), False
gbm = lgb.train(params,
lgb_train,
num_boost_round=10,
init_model=gbm,
fobj=loglikelood,
feval=binary_error,
valid_sets=lgb_eval)
print('用自定义的损失函数与评估标准完成第40-50轮...')
作者：小菜鸡一号
