FLAML多输出回归预测的写法示例，及整体评估和分输出变量评估

已于 2024-10-25 18:01:12 修改
阅读量854 收藏 16
点赞数 7
分类专栏： ML与AI 文章标签： FLAML AutoML 机器学习
于 2024-10-24 18:28:40 首次发布
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.youkuaiyun.com/piaopiaolanghua/article/details/143216407
版权

【摘要】

1、我们可以结合sklearn.MultiOutputRegressor和flaml.AutoML来对多输出变量回归模型进行训练。它启动一次，将为每个目标执行自动化机器学习，这是其方便的地方所在。

2、创建随机输入数据代码如下，输入变量为5个，输出变量为3个。

from sklearn.datasets import make_regression

# 目标变量数量n_targets, 特征数量n_features
n_targets=3
X, y = make_regression(n_targets=n_targets, n_features=5)

【官方参考】

AutoML - Regression | FLAML

【关键输出】

# python test-flaml-mo.py
[flaml.automl.logger: 10-25 00:34:32] {1728} INFO - task = regression
[flaml.automl.logger: 10-25 00:34:32] {1739} INFO - Evaluation method: cv
[flaml.automl.logger: 10-25 00:34:32] {1838} INFO - Minimizing error metric: 1-r2
[flaml.automl.logger: 10-25 00:34:32] {1955} INFO - List of ML learners in AutoML Run: ['lgbm', 'rf', 'xgboost', 'extra_tree', 'xgb_limitdepth', 'sgd']
[flaml.automl.logger: 10-25 00:34:32] {2258} INFO - iteration 0, current learner lgbm
[flaml.automl.logger: 10-25 00:34:32] {2393} INFO - Estimated sufficient time budget=207s. Estimated necessary time budget=1s.
[flaml.automl.logger: 10-25 00:34:32] {2442} INFO -  at 0.0s,   estimator lgbm's best error=0.8370,     best estimator lgbm's best error=0.8370
[flaml.automl.logger: 10-25 00:34:32] {2258} INFO - iteration 1, current learner lgbm
[flaml.automl.logger: 10-25 00:34:32] {2442} INFO -  at 0.0s,   estimator lgbm's best error=0.8370,     best estimator lgbm's best error=0.8370
[flaml.automl.logger: 10-25 00:34:32] {2258} INFO - iteration 2, current learner lgbm
[flaml.automl.logger: 10-25 00:34:32] {2442} INFO -  at 0.1s,   estimator lgbm's best error=0.5070,     best estimator lgbm's best error=0.5070

...

[flaml.automl.logger: 10-25 00:34:44] {2258} INFO - iteration 125, current learner sgd
[flaml.automl.logger: 10-25 00:34:44] {2442} INFO -  at 4.0s,   estimator sgd's best error=0.0716,      best estimator sgd's best error=0.0716
[flaml.automl.logger: 10-25 00:34:44] {2685} INFO - retrain sgd for 0.0s
[flaml.automl.logger: 10-25 00:34:44] {2688} INFO - retrained model: SGDRegressor(alpha=9.820909968561154e-05, learning_rate='optimal',
             loss='epsilon_insensitive', penalty=None, tol=0.0001)
[flaml.automl.logger: 10-25 00:34:44] {1985} INFO - fit succeeded
[flaml.automl.logger: 10-25 00:34:44] {1986} INFO - Time taken to find the best model: 3.348128318786621


first result: [[-68.01898084 114.18330859  76.12119675]]


R2 Score: 0.8804327929345547
(MSE )Mean Squared Error     : 2064.864234989927
(RMSE)Root Mean Squared Error: 45.440777226956925
(MAE )Mean Absolute Error    : 26.20185200402915


R2 Score 0: 0.8857770110995447
(MSE )Mean Squared Error 0     : 637.5601445655168
(RMSE)Root Mean Squared Error 0: 25.249953357689925
(MAE )Mean Absolute Error 0    : 17.936946049090224


R2 Score 1: 0.8838947860589346
(MSE )Mean Squared Error 1     : 3615.804943742679
(RMSE)Root Mean Squared Error 1: 60.131563622964926
(MAE )Mean Absolute Error 1    : 35.68283164689059


R2 Score 2: 0.8716265816451845
(MSE )Mean Squared Error 2     : 1941.2276166615861
(RMSE)Root Mean Squared Error 2: 44.05936468744853
(MAE )Mean Absolute Error 2    : 24.985778316106625

 【带整体评估和分输出变量（特征）评估例子】

from flaml import AutoML
from sklearn.datasets import make_regression
from sklearn.model_selection import train_test_split
from sklearn.multioutput import MultiOutputRegressor
from sklearn.metrics import mean_squared_error, r2_score, mean_absolute_error
import numpy as np
import math

# 目标变量数量n_targets, 特征数量n_features
n_targets=3
X, y = make_regression(n_targets=n_targets, n_features=5)

# split into train and test data
X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.30, random_state=42)

first_row = X_test[0,:]
first_col = X_test[:,0]

# train the model
model = MultiOutputRegressor(AutoML(task="regression", time_budget=4))
model.fit(X_train, y_train)

# predict
y_pred = model.predict(X_test)

# 单个预测
npa = np.array([first_row])
print('first result:', model.predict(npa))

# 总体评估
r2 = r2_score(y_test, y_pred)
mse = mean_squared_error(y_test, y_pred)
rmse = math.sqrt(mse)
mae = mean_absolute_error(y_test, y_pred)
print(f"R2 Score: {r2}")
print(f"(MSE )Mean Squared Error     : {mse}")
print(f"(RMSE)Root Mean Squared Error: {rmse}")
print(f"(MAE )Mean Absolute Error    : {mae}")
print()


# 分列总体评估
for i in range(n_targets):
    yi_test=y_test[:,i]
    yi_pred=y_pred[:,i]

    r2 = r2_score(yi_test, yi_pred)
    mse = mean_squared_error(yi_test, yi_pred)
    rmse = math.sqrt(mse)
    mae = mean_absolute_error(yi_test, yi_pred)

    print(f"R2 Score {i}: {r2}")
    print(f"(MSE )Mean Squared Error {i}     : {mse}")
    print(f"(RMSE)Root Mean Squared Error {i}: {rmse}")
    print(f"(MAE )Mean Absolute Error {i}    : {mae}")
    print()

【完整输出】

# python test-flaml-mo.py
[flaml.automl.logger: 10-25 00:34:32] {1728} INFO - task = regression
[flaml.automl.logger: 10-25 00:34:32] {1739} INFO - Evaluation method: cv
[flaml.automl.logger: 10-25 00:34:32] {1838} INFO - Minimizing error metric: 1-r2
[flaml.automl.logger: 10-25 00:34:32] {1955} INFO - List of ML learners in AutoML Run: ['lgbm', 'rf', 'xgboost', 'extra_tree', 'xgb_limitdepth', 'sgd']
[flaml.automl.logger: 10-25 00:34:32] {2258} INFO - iteration 0, current learner lgbm
[flaml.automl.logger: 10-25 00:34:32] {2393} INFO - Estimated sufficient time budget=207s. Estimated necessary time budget=1s.
[flaml.automl.logger: 10-25 00:34:32] {2442} INFO -  at 0.0s,   estimator lgbm's best error=0.8370,     best estimator lgbm's best error=0.8370
[flaml.automl.logger: 10-25 00:34:32] {2258} INFO - iteration 1, current learner lgbm
[flaml.automl.logger: 10-25 00:34:32] {2442} INFO -  at 0.0s,   estimator lgbm's best error=0.8370,     best estimator lgbm's best error=0.8370
[flaml.automl.logger: 10-25 00:34:32] {2258} INFO - iteration 2, current learner lgbm
[flaml.automl.logger: 10-25 00:34:32] {2442} INFO -  at 0.1s,   estimator lgbm's best error=0.5070,     best estimator lgbm's best error=0.5070
[flaml.automl.logger: 10-25 00:34:32] {2258} INFO - iteration 3, current learner sgd
[flaml.automl.logger: 10-25 00:34:32] {2442} INFO -  at 0.1s,   estimator sgd's best error=0.8813,      best estimator lgbm's best error=0.5070
[flaml.automl.logger: 10-25 00:34:32] {2258} INFO - iteration 4, current learner lgbm
[flaml.automl.logger: 10-25 00:34:32] {2442} INFO -  at 0.1s,   estimator lgbm's best error=0.3054,     best estimator lgbm's best error=0.3054
[flaml.automl.logger: 10-25 00:34:32] {2258} INFO - iteration 5, current learner lgbm
[flaml.automl.logger: 10-25 00:34:32] {2442} INFO -  at 0.1s,   estimator lgbm's best error=0.3054,     best estimator lgbm's best error=0.3054
[flaml.automl.logger: 10-25 00:34:32] {2258} INFO - iteration 6, current learner lgbm
[flaml.automl.logger: 10-25 00:34:32] {2442} INFO -  at 0.1s,   estimator lgbm's best error=0.2864,     best estimator lgbm's best error=0.2864
[flaml.automl.logger: 10-25 00:34:32] {2258} INFO - iteration 7, current learner xgboost
[flaml.automl.logger: 10-25 00:34:32] {2442} INFO -  at 0.2s,   estimator xgboost's best error=0.7272,  best estimator lgbm's best error=0.2864
[flaml.automl.logger: 10-25 00:34:32] {2258} INFO - iteration 8, current learner lgbm
[flaml.automl.logger: 10-25 00:34:32] {2442} INFO -  at 0.2s,   estimator lgbm's best error=0.2864,     best estimator lgbm's best error=0.2864
[flaml.automl.logger: 10-25 00:34:32] {2258} INFO - iteration 9, current learner lgbm
[flaml.automl.logger: 10-25 00:34:32] {2442} INFO -  at 0.2s,   estimator lgbm's best error=0.2864,     best estimator lgbm's best error=0.2864
[flaml.automl.logger: 10-25 00:34:32] {2258} INFO - iteration 10, current learner lgbm
[flaml.automl.logger: 10-25 00:34:33] {2442} INFO -  at 0.2s,   estimator lgbm's best error=0.2021,     best estimator lgbm's best error=0.2021
[flaml.automl.logger: 10-25 00:34:33] {2258} INFO - iteration 11, current learner extra_tree
[flaml.automl.logger: 10-25 00:34:33] {2442} INFO -  at 0.4s,   estimator extra_tree's best error=0.6197,       best estimator lgbm's best error=0.2021
[flaml.automl.logger: 10-25 00:34:33] {2258} INFO - iteration 12, current learner rf
[flaml.automl.logger: 10-25 00:34:33] {2442} INFO -  at 0.6s,   estimator rf's best error=0.6211,       best estimator lgbm's best error=0.2021
[flaml.automl.logger: 10-25 00:34:33] {2258} INFO - iteration 13, current learner xgboost
[flaml.automl.logger: 10-25 00:34:33] {2442} INFO -  at 0.6s,   estimator xgboost's best error=0.7272,  best estimator lgbm's best error=0.2021
[flaml.automl.logger: 10-25 00:34:33] {2258} INFO - iteration 14, current learner xgboost
[flaml.automl.logger: 10-25 00:34:33] {2442} INFO -  at 0.6s,   estimator xgboost's best error=0.4619,  best estimator lgbm's best error=0.2021
[flaml.automl.logger: 10-25 00:34:33] {2258} INFO - iteration 15, current learner xgboost
[flaml.automl.logger: 10-25 00:34:33] {2442} INFO -  at 0.7s,   estimator xgboost's best error=0.4619,  best estimator lgbm's best error=0.2021
[flaml.automl.logger: 10-25 00:34:33] {2258} INFO - iteration 16, current learner sgd
[flaml.automl.logger: 10-25 00:34:33] {2442} INFO -  at 0.7s,   estimator sgd's best error=0.8813,      best estimator lgbm's best error=0.2021
[flaml.automl.logger: 10-25 00:34:33] {2258} INFO - iteration 17, current learner lgbm
[flaml.automl.logger: 10-25 00:34:33] {2442} INFO -  at 0.7s,   estimator lgbm's best error=0.2021,     best estimator lgbm's best error=0.2021
[flaml.automl.logger: 10-25 00:34:33] {2258} INFO - iteration 18, current learner sgd
[flaml.automl.logger: 10-25 00:34:33] {2442} INFO -  at 0.7s,   estimator sgd's best error=0.2627,      best estimator lgbm's best error=0.2021
[flaml.automl.logger: 10-25 00:34:33] {2258} INFO - iteration 19, current learner lgbm
[flaml.automl.logger: 10-25 00:34:33] {2442} INFO -  at 0.8s,   estimator lgbm's best error=0.2021,     best estimator lgbm's best error=0.2021
[flaml.automl.logger: 10-25 00:34:33] {2258} INFO - iteration 20, current learner sgd
[flaml.automl.logger: 10-25 00:34:33] {2442} INFO -  at 0.8s,   estimator sgd's best error=0.2627,      best estimator lgbm's best error=0.2021
[flaml.automl.logger: 10-25 00:34:33] {2258} INFO - iteration 21, current learner xgboost
[flaml.automl.logger: 10-25 00:34:33] {2442} INFO -  at 0.8s,   estimator xgboost's best error=0.4619,  best estimator lgbm's best error=0.2021
[flaml.automl.logger: 10-25 00:34:33] {2258} INFO - iteration 22, current learner extra_tree
[flaml.automl.logger: 10-25 00:34:33] {2442} INFO -  at 1.0s,   estimator extra_tree's best error=0.4389,       best estimator lgbm's best error=0.2021
[flaml.automl.logger: 10-25 00:34:33] {2258} INFO - iteration 23, current learner lgbm
[flaml.automl.logger: 10-25 00:34:33] {2442} INFO -  at 1.0s,   estimator lgbm's best error=0.2021,     best estimator lgbm's best error=0.2021
[flaml.automl.logger: 10-25 00:34:33] {2258} INFO - iteration 24, current learner lgbm
[flaml.automl.logger: 10-25 00:34:33] {2442} INFO -  at 1.0s,   estimator lgbm's best error=0.2021,     best estimator lgbm's best error=0.2021
[flaml.automl.logger: 10-25 00:34:33] {2258} INFO - iteration 25, current learner sgd
[flaml.automl.logger: 10-25 00:34:33] {2442} INFO -  at 1.0s,   estimator sgd's best error=0.2627,      best estimator lgbm's best error=0.2021
[flaml.automl.logger: 10-25 00:34:33] {2258} INFO - iteration 26, current learner sgd
[flaml.automl.logger: 10-25 00:34:33] {2442} INFO -  at 1.0s,   estimator sgd's best error=0.2627,      best estimator lgbm's best error=0.2021
[flaml.automl.logger: 10-25 00:34:33] {2258} INFO - iteration 27, current learner rf
[flaml.automl.logger: 10-25 00:34:33] {2442} INFO -  at 1.2s,   estimator rf's best error=0.4799,       best estimator lgbm's best error=0.2021
[flaml.automl.logger: 10-25 00:34:33] {2258} INFO - iteration 28, current learner sgd
[flaml.automl.logger: 10-25 00:34:33] {2442} INFO -  at 1.2s,   estimator sgd's best error=0.1883,      best estimator sgd's best error=0.1883
[flaml.automl.logger: 10-25 00:34:33] {2258} INFO - iteration 29, current learner sgd
[flaml.automl.logger: 10-25 00:34:33] {2442} INFO -  at 1.2s,   estimator sgd's best error=0.1532,      best estimator sgd's best error=0.1532
[flaml.automl.logger: 10-25 00:34:33] {2258} INFO - iteration 30, current learner sgd
[flaml.automl.logger: 10-25 00:34:34] {2442} INFO -  at 1.2s,   estimator sgd's best error=0.1532,      best estimator sgd's best error=0.1532
[flaml.automl.logger: 10-25 00:34:34] {2258} INFO - iteration 31, current learner xgboost
[flaml.automl.logger: 10-25 00:34:34] {2442} INFO -  at 1.3s,   estimator xgboost's best error=0.4619,  best estimator sgd's best error=0.1532
[flaml.automl.logger: 10-25 00:34:34] {2258} INFO - iteration 32, current learner lgbm
[flaml.automl.logger: 10-25 00:34:34] {2442} INFO -  at 1.3s,   estimator lgbm's best error=0.2021,     best estimator sgd's best error=0.1532
[flaml.automl.logger: 10-25 00:34:34] {2258} INFO - iteration 33, current learner sgd
[flaml.automl.logger: 10-25 00:34:34] {2442} INFO -  at 1.3s,   estimator sgd's best error=0.1300,      best estimator sgd's best error=0.1300
[flaml.automl.logger: 10-25 00:34:34] {2258} INFO - iteration 34, current learner sgd
[flaml.automl.logger: 10-25 00:34:34] {2442} INFO -  at 1.3s,   estimator sgd's best error=0.1300,      best estimator sgd's best error=0.1300
[flaml.automl.logger: 10-25 00:34:34] {2258} INFO - iteration 35, current learner xgboost
[flaml.automl.logger: 10-25 00:34:34] {2442} INFO -  at 1.4s,   estimator xgboost's best error=0.3604,  best estimator sgd's best error=0.1300
[flaml.automl.logger: 10-25 00:34:34] {2258} INFO - iteration 36, current learner sgd
[flaml.automl.logger: 10-25 00:34:34] {2442} INFO -  at 1.4s,   estimator sgd's best error=0.1300,      best estimator sgd's best error=0.1300
[flaml.automl.logger: 10-25 00:34:34] {2258} INFO - iteration 37, current learner sgd
[flaml.automl.logger: 10-25 00:34:34] {2442} INFO -  at 1.4s,   estimator sgd's best error=0.1300,      best estimator sgd's best error=0.1300
[flaml.automl.logger: 10-25 00:34:34] {2258} INFO - iteration 38, current learner sgd
[flaml.automl.logger: 10-25 00:34:34] {2442} INFO -  at 1.4s,   estimator sgd's best error=0.1300,      best estimator sgd's best error=0.1300
[flaml.automl.logger: 10-25 00:34:34] {2258} INFO - iteration 39, current learner sgd
[flaml.automl.logger: 10-25 00:34:34] {2442} INFO -  at 1.4s,   estimator sgd's best error=0.1300,      best estimator sgd's best error=0.1300
[flaml.automl.logger: 10-25 00:34:34] {2258} INFO - iteration 40, current learner extra_tree
[flaml.automl.logger: 10-25 00:34:34] {2442} INFO -  at 1.6s,   estimator extra_tree's best error=0.4389,       best estimator sgd's best error=0.1300
[flaml.automl.logger: 10-25 00:34:34] {2258} INFO - iteration 41, current learner sgd
[flaml.automl.logger: 10-25 00:34:34] {2442} INFO -  at 1.6s,   estimator sgd's best error=0.1300,      best estimator sgd's best error=0.1300
[flaml.automl.logger: 10-25 00:34:34] {2258} INFO - iteration 42, current learner sgd
[flaml.automl.logger: 10-25 00:34:34] {2442} INFO -  at 1.6s,   estimator sgd's best error=0.1300,      best estimator sgd's best error=0.1300
[flaml.automl.logger: 10-25 00:34:34] {2258} INFO - iteration 43, current learner lgbm
[flaml.automl.logger: 10-25 00:34:34] {2442} INFO -  at 1.6s,   estimator lgbm's best error=0.2021,     best estimator sgd's best error=0.1300
[flaml.automl.logger: 10-25 00:34:34] {2258} INFO - iteration 44, current learner sgd
[flaml.automl.logger: 10-25 00:34:34] {2442} INFO -  at 1.6s,   estimator sgd's best error=0.1300,      best estimator sgd's best error=0.1300
[flaml.automl.logger: 10-25 00:34:34] {2258} INFO - iteration 45, current learner sgd
[flaml.automl.logger: 10-25 00:34:34] {2442} INFO -  at 1.7s,   estimator sgd's best error=0.1300,      best estimator sgd's best error=0.1300
[flaml.automl.logger: 10-25 00:34:34] {2258} INFO - iteration 46, current learner sgd
[flaml.automl.logger: 10-25 00:34:34] {2442} INFO -  at 1.7s,   estimator sgd's best error=0.1300,      best estimator sgd's best error=0.1300
[flaml.automl.logger: 10-25 00:34:34] {2258} INFO - iteration 47, current learner rf
[flaml.automl.logger: 10-25 00:34:34] {2442} INFO -  at 1.8s,   estimator rf's best error=0.4799,       best estimator sgd's best error=0.1300
[flaml.automl.logger: 10-25 00:34:34] {2258} INFO - iteration 48, current learner sgd
[flaml.automl.logger: 10-25 00:34:34] {2442} INFO -  at 1.8s,   estimator sgd's best error=0.1248,      best estimator sgd's best error=0.1248
[flaml.autom
最低0.47元/天 解锁文章
评论
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包

打赏作者

piaopiaolanghua

感谢老板鼓励~

¥1 ¥2 ¥4 ¥6 ¥10 ¥20
扫码支付：¥1
获取中
扫码支付

您的余额不足，请更换扫码支付或充值

打赏作者

实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值