lgb多分类参数设置

版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
本文链接：https://blog.csdn.net/weixin_41089007/article/details/90510248

数据

train_x, test_x, train_y, test_y = train_test_split(data, target, shuffle = True, random_state = 2019)
X_train = train_x.values
X_test = test_x.values
y_train = train_y.values
y_test = test_y.values

方式一：直接跑

params = {  
    'boosting_type': 'gbdt',  
    'objective': 'multiclass',  
    'num_class': 7,  
    'metric': 'multi_error',  
    'num_leaves': 120,  
    'min_data_in_leaf': 100,  
    'learning_rate': 0.06,  
    'feature_fraction': 0.8,  
    'bagging_fraction': 0.8,  
    'bagging_freq': 5,  
    'lambda_l1': 0.4,  
    'lambda_l2': 0.5,  
    'min_gain_to_split': 0.2,  
    'verbose': -1, 
}  
print('Training...')
trn_data = lgb.Dataset(X_train, y_train)
val_data = lgb.Dataset(X_test, y_test)
clf = lgb.train(params, 
                trn_data, 
                num_boost_round = 1000,
                valid_sets = [trn_data,val_data], 
                verbose_eval = 100, 
                early_stopping_rounds = 100)
print('Predicting...')
y_prob = clf.predict(X_test, num_iteration=clf.best_iteration)
y_pred = [list(x).index(max(x)) for x in y_prob]
print("AUC score: {:<8.5f}".format(metrics.accuracy_score(y_pred, test_y)))

方式二：加入交叉验证

param = { 
    'boosting_type': 'gbdt',  
    'objective': 'multiclass',  
    'num_class': 7,  
    'metric': 'multi_error',  
    'num_leaves': 300,  
    'min_data_in_leaf': 500,  
    'learning_rate': 0.01,  
    'feature_fraction': 0.8,  
    'bagging_fraction': 0.8,  
    'bagging_freq': 5,  
    'lambda_l1': 0.4,  
    'lambda_l2': 0.5,  
    'min_gain_to_split': 0.2,  
    'verbose': -1,
    'num_threads':4,
}
 
# 五折交叉验证
folds = KFold(n_splits=5, shuffle=False, random_state=2019)
oof = np.zeros([len(X_train),7])
predictions = np.zeros([len(X_test),7])
 
for fold_, (trn_idx, val_idx) in enumerate(folds.split(X_train, y_train)):
    print("fold n°{}".format(fold_+1))
    trn_data = lgb.Dataset(X_train[trn_idx], y_train[trn_idx])
    val_data = lgb.Dataset(X_train[val_idx], y_train[val_idx])
 
    num_round = 1000
    clf = lgb.train(param, 
                    trn_data, 
                    num_round, 
                    valid_sets = [trn_data, val_data], 
                    verbose_eval = 100, 
                    early_stopping_rounds = 100)
    #oof[val_idx] = clf.predict(X_train[val_idx], num_iteration=clf.best_iteration)    
    predictions += clf.predict(X_test, num_iteration=clf.best_iteration) / folds.n_splits
    #print(predictions)

————————————————
版权声明：本文为CSDN博主「睡熊猛醒」的原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接及本声明。
原文链接：https://blog.csdn.net/weixin_41089007/article/details/90510248