EasyTableML is an automated machine learning framework suitable for tabular data, which is very convenient to use. You only need 1-2 lines of code to train an excellent machine learning model.
Step one : Clone this repository using git clone or download this repository
Step two : Using pip install -r requirements.txt to install dependencies
Step three : Feel free to use 😊
You can train a machine learning model with your data in just four simple steps:
Step one : Import module
import EasyTableML
Step two : Initialize the EasyTableMLRegression object
easyml=EasyTableML.EasyTableMLRegression()
Step three (optional) : Extract automatic features
x_train, x_test, y_train=easyml.auto_features(x_train, x_test, y_train, n_jobs="The number of your cpu cores")
Note: x_train x_test, y_train Must be pandas.DataFrame
Step Four : Train model
model=easyml.fit(x_train,y_train,n_jobs="The number of your cpu cores")
Use easyml.predict(model, x_test) to make predictions about data
Function: Automatic extraction of training features, batch normalization of data, improve the accuracy of subsequent training process
Parameters
x_train:Training features. The type must be pandas.DataFrame. It is recommended to encode the text and remove null values before input.
x_test:Test features (predicted features). The type must be pandas.DataFram. It is recommended to encode the text and remove null values before input.
y_train:Training labels. The type must be pandas.DataFrame.
features_number:The number of features to be expanded in automatic feature extraction. The higher the value, the higher the model accuracy, but the slower the training speed. The default value is 20.
n_jobs:The number of resources used in automatic feature extraction. It is recommended to set to the number of CPU cores. The default value is 1.
Return
Transformed x_train, x_test, y_train, type : numpy.array
Function: Gets the default base learning model.
Parameters
None.
Return
Model dictionary {'model_name':model_object}
train(model_list,x_train,y_train,train_type,meta_model=None,cv=5,auto_custom_parameters=None,n_jobs=1,auto_scoring='neg_mean_absolute_error',details=1,auto_parameter=False):
Function: Training model.
Parameters
model_list:Model dictionary {'model_name':model_object}.
x_train:Training features. The type must be numpy.array.
y_train:Training labels. The type must be numpy.array. The shape must be (labels_number,)
train_type:'base' or 'meta'. 'base' trains the base model and 'meta' trains the metamodel.
meta_model:If it is not None, the default metamodel is not used for training. Use a custom metamodel. auto_parameter must also be set to False to take effect. The default value is None
cv:The number of splits when cross-verifying. The default value is 5
auto_custom_parameters:When auto_parameter is set to True, you can customize the parameter search range by specifying this parameter. The parameter input format is {'model_name':{'parameter_name',[parameter_range]}}. The default value is None
n_jobs:Resources used during training. It is recommended to set to the number of CPU cores. The default value is 1
auto_scoring:Model evaluation index in automatic parameter search. This parameter is invalid when auto_parameter is set to False. For the value of this parameter, see: https://scikit-learn.org/stable/modules/model_evaluation.html. The default value is neg_mean_absolute_error
details:The level of detail of the output information.The larger the value, the more detailed the output information. The default value is 1
auto_parameter:True or False. Whether to use automatic parameter search. If it is set to False, the automatic parameter search is not performed and the model provided in model_list or meta_model is used directly for training. If set to True, the optimal parameters for the model in model_list are searched automatically.
Return
If train_type is 'base', return the trained model dictionary {'model_name':model_object}
If train_type is 'meta', return metamodel object.
Function: Load the locally saved base models.
Parameters
model_list:Model dictionary {'model_name':model_object}.
is_auto_parameter:True or False. If True, loads the model saved when using automatic parameter search; If False, load the model that was saved when automatic parameter search was not used. The default value is True.
Return
Returns model dictionary with loaded data {'model_name':model_object}
Function: Load the locally saved meta model.
Parameters
is_auto_parameter:True or False. If True, loads the model saved when using automatic parameter search; If False, load the model that was saved when automatic parameter search was not used. The default value is True.
Return
Returns metamodel with loaded data.
Function: Evaluate the training effect of the base models and output MAE, RMSE and R2 scores of the base model.
Parameters
model_list:Model dictionary {'model_name':model_object}.
x_train:Training features. The type must be numpy.array.
y_train:Training labels. The type must be numpy.array. The shape must be (labels_number,)
x_valid:Test or validation features. The type must be numpy.array.
y_valid:Test or validation labels. The type must be numpy.array. The shape must be (labels_number,)
details:The level of detail of the output information.The larger the value, the more detailed the output information. The default value is 0
Return
None.
Function: Evaluate the training effect of the metamodel and output MAE, RMSE and R2 scores of the base model.
Parameters
meta_model:Metamodel object.
x_train:Training features. The type must be numpy.array.
y_train:Training labels. The type must be numpy.array. The shape must be (labels_number,)
x_valid:Test or validation features. The type must be numpy.array.
y_valid:Test or validation labels. The type must be numpy.array. The shape must be (labels_number,)
details:The level of detail of the output information.The larger the value, the more detailed the output information. The default value is 0
Return
None.
Function:Predict the value of the label.
Parameters
model:A model object.
x_pred:Features to be predicted. The type must be numpy.array.
Return
Predicted results. The type is numpy.array.
Function:One line of code gets the training model.
Parameters
x_train:Training features. The type must be numpy.array.
y_train:Training labels. The type must be numpy.array. The shape must be (labels_number,)
auto_scoring:Model evaluation index in automatic parameter search. This parameter is invalid when auto_parameter is set to False. For the value of this parameter, see: https://scikit-learn.org/stable/modules/model_evaluation.html. The default value is neg_mean_absolute_error
cv:The number of splits when cross-verifying. The default value is 5
n_jobs:Resources used during training. It is recommended to set to the number of CPU cores. The default value is 1
details:The level of detail of the output information.The larger the value, the more detailed the output information. The default value is 1