AutoML框架
相对于其他 AutoML工具来说， Auto_ml 是从0->1的开创性的工作，其附加功能很多，比如feature response，data clean，ml analysis，feature learning ，deep learning，Categorical Ensembling（类别组合）等。但是对于auto_ml的核心——model selection，hyper parameter optimation等，它仅仅做的很基础——使用SGCV进行网格搜索，而在不选择model selection时，model是默认给定的（默认GradientBoostingRegressor for regressor &GradientBoostingClassifier for classifier）。 总之，auto_ml做的很基础很全面，根据阅读源代码熟悉 autoML的Pipeline，适合auto_ml新手进行探索。
特性
使整个机器学习过程自动化，使其非常易于用于分析和获取生产中的实时预测。该项目自动执行：
- Analytics (pass in data, and auto_ml will tell you the relationship of each variable to what it is you’re trying to predict).
- Feature Engineering (particularly around dates, and soon, NLP).
- Robust Scaling (turning all values into their scaled versions between the range of 0 and 1, in a way that is robust to outliers, and works with sparse matrices).
- Feature Selection (picking only the features that actually prove useful).
- Data formatting (turning a list of dictionaries into a sparse matrix, one-hot encoding categorical variables, taking the natural log of y for regression problems).
- Model Selection (which model works best for your problem).
- Hyperparameter Optimization (what hyperparameters work best for that model).
- Ensembling Subpredictors (automatically training up models to predict smaller problems within the meta problem).
- Ensembling Weak Estimators (automatically training up weak models on the larger problem itself, to inform the meta-estimator’s decision).
- Big Data (feed it lots of data).
- Unicorns (you could conceivably train it to predict what is a unicorn and what is not).
- Hugs (this makes it much easier to do your job, hopefully leaving you more time to hug those those you care about).
GetStart
Installation
auto_ml模块可直接使用pip命令进行安装，目前版本为auto_ml 2.9.10
python
pip install auto_ml第三方软件包
- TensorFlow & Keras, XGBoost, LightGBM
auto_ml集成了所有这些很棒的库，包括 DeepLearningClassifier 、 DeepLearningRegressor - XGBClassifier、 XGBRegressor - LGBMClassifer、 LGBMRegressor - CatBoostClassifier和 CatBoostRegressor等等，这些项目对单个预测的预测时间都在1毫秒范围内，并且能够序列化到磁盘，并在训练后加载到新的环境中。auto_ml在没有安装它们的情况下可以很好地运行。只需为model_names传递其中之一即可实现调用。
python
ml_predictor.train(data, model_names=['DeepLearningClassifier'])使用示例
1、导入auto_ml中的包Predictor和get_boston_dataset
2、通过get_boston_dataset获取训练集和测试集
3、告诉机器数据输出列和不是纯数字的列
4、建立模型并训练模型
5、模型训练效果评估
6、保存模型
7、使用模型进行预测
python
from auto_ml import Predictor
from auto_ml.utils import get_boston_dataset
df_train, df_test = get_boston_dataset()
column_descriptions = {
'MEDV': 'output'
, 'CHAS': 'categorical'
}
ml_predictor = Predictor(type_of_estimator='regressor', column_descriptions=column_descriptions)
ml_predictor.train(df_train)
test_score = ml_predictor.score(df_test, df_test.MEDV)
file_name = ml_predictor.save()
trained_model = load_ml_model(file_name)evironments)
predictions = trained_model.predict(df_test)
print(predictions)运行结束
python
+----+----------------+--------------+----------+-------------------+-------------------+-----------+-----------+-----------+-----------+
| | Feature Name | Importance | Delta | FR_Decrementing | FR_Incrementing | FRD_abs | FRI_abs | FRD_MAD | FRI_MAD |
|----+----------------+--------------+----------+-------------------+-------------------+-----------+-----------+-----------+-----------|
| 12 | CHAS | 7 | nan | nan | nan | nan | nan | nan | nan |
| 1 | ZN | 12 | 11.5619 | -0.0062 | 0.0380 | 0.0239 | 0.0474 | 0.0000 | 0.0000 |
| 7 | RAD | 17 | 4.2895 | -0.1370 | 0.1163 | 0.1588 | 0.1673 | 0.1175 | 0.1035 |
| 2 | INDUS | 42 | 3.4430 | 0.1632 | -0.1283 | 0.2332 | 0.2536 | 0.0278 | 0.0785 |
| 8 | TAX | 50 | 82.9834 | 0.7771 | -0.3183 | 0.8333 | 0.4623 | 0.3680 | 0.1181 |
| 9 | PTRATIO | 59 | 1.1130 | 0.6408 | -0.4830 | 0.6931 | 0.5645 | 0.4029 | 0.2925 |
| 3 | NOX | 79 | 0.0588 | 0.4932 | -0.1203 | 0.8052 | 0.6728 | 0.1935 | 0.2717 |
| 0 | CRIM | 142 | 4.4320 | -0.6586 | 1.3412 | 0.9985 | 1.7242 | 0.9479 | 1.6564 |
| 5 | AGE | 143 | 13.9801 | 0.4331 | -0.4677 | 0.6308 | 0.6133 | 0.5353 | 0.4838 |
| 6 | DIS | 176 | 1.0643 | 1.7858 | -0.3931 | 1.8339 | 0.8874 | 1.0996 | 0.3497 |
| 10 | B | 179 | 45.7266 | -0.7798 | 0.1640 | 1.1877 | 0.4785 | 1.0904 | 0.1836 |
| 4 | RM | 197 | 0.3543 | -1.1633 | 1.6018 | 1.3489 | 1.7951 | 0.9774 | 1.2344 |
| 11 | LSTAT | 213 | 3.5508 | 2.5125 | -1.4640 | 2.6368 | 1.5911 | 1.7992 | 1.1952 |
+----+----------------+--------------+----------+-------------------+-------------------+-----------+-----------+-----------+-----------+python
***********************************************
Advanced scoring metrics for the trained regression model on this particular dataset:
Here is the overall RMSE for these predictions:
2.7608924254130724
Here is the average of the predictions:
21.350718828184544
Here is the average actual value on this validation set:
21.488235294117654
Here is the median prediction:
20.70223186516366
Here is the median actual value:
20.15
Here is the mean absolute error:
2.055851103106312
Here is the median absolute error (robust to outliers):
1.625601367533859
Here is the explained variance:
0.8963149604640271
Here is the R-squared value:
0.8960570877948729
Count of positive differences (prediction > actual):
48
Count of negative differences:
54
Average positive difference:
2.0382305519965307
Average negative difference:
-2.071513815203894
***********************************************