LGB(机器学习模板)

import lightgbm as lgb

import pandas as pd

import numpy as np

import pickle

from sklearn.metrics import roc_auc_score

from sklearn.model_selection import train_test_split



print("Loading Data ... ")



# 导入数据

train_x, train_y, test_x = load_data()



# 用sklearn.cross_validation进行训练数据集划分，这里训练集和交叉验证集比例为7：3，可以自己根据需要设置

X, val_X, y, val_y = train_test_split(

    train_x,

    train_y,

    test_size=0.05,

    random_state=1,

    stratify=train_y ## 这里保证分割后y的比例分布与原数据一致

)



X_train = X

y_train = y

X_test = val_X

y_test = val_y





# create dataset for lightgbm

lgb_train = lgb.Dataset(X_train, y_train)

lgb_eval = lgb.Dataset(X_test, y_test, reference=lgb_train)

# specify your configurations as a dict

params = {

    'boosting_type': 'gbdt',

    'objective': 'binary',

    'metric': {'binary_logloss', 'auc'},

    'num_leaves': 5,

    'max_depth': 6,

    'min_data_in_leaf': 450,

    'learning_rate': 0.1,

    'feature_fraction': 0.9,

    'bagging_fraction': 0.95,

    'bagging_freq': 5,

    'lambda_l1': 1,  

    'lambda_l2': 0.001,  # 越小l2正则程度越高

    'min_gain_to_split': 0.2,

    'verbose': 5,

    'is_unbalance': True

}



# train

print('Start training...')

gbm = lgb.train(params,

                lgb_train,

                num_boost_round=10000,

                valid_sets=lgb_eval,

                early_stopping_rounds=500)



print('Start predicting...')



preds = gbm.predict(test_x, num_iteration=gbm.best_iteration)  # 输出的是概率结果



# 导出结果

threshold = 0.5

for pred in preds:

    result = 1 if pred > threshold else 0



# 导出特征重要性

importance = gbm.feature_importance()

names = gbm.feature_name()

with open('./feature_importance.txt', 'w+') as file:

    for index, im in enumerate(importance):

        string = names[index] + ', ' + str(im) + '\n'

        file.write(string)