How to import a class object from a list of objects?-CodePudding

Using the below code:

from sklearn.utils import all_estimators
from sklearn import base

# Print all regressors
estimators = all_estimators(type_filter="regressor")
for name in estimators:
    print(name[0], name[1])

...renders this response:

ARDRegression <class 'sklearn.linear_model._bayes.ARDRegression'>
AdaBoostRegressor <class 'sklearn.ensemble._weight_boosting.AdaBoostRegressor'>
BaggingRegressor <class 'sklearn.ensemble._bagging.BaggingRegressor'>
BayesianRidge <class 'sklearn.linear_model._bayes.BayesianRidge'>
CCA <class 'sklearn.cross_decomposition._pls.CCA'>
DecisionTreeRegressor <class 'sklearn.tree._classes.DecisionTreeRegressor'>
DummyRegressor <class 'sklearn.dummy.DummyRegressor'>
ElasticNet <class 'sklearn.linear_model._coordinate_descent.ElasticNet'>
ElasticNetCV <class 'sklearn.linear_model._coordinate_descent.ElasticNetCV'>
ExtraTreeRegressor <class 'sklearn.tree._classes.ExtraTreeRegressor'>
ExtraTreesRegressor <class 'sklearn.ensemble._forest.ExtraTreesRegressor'>
GammaRegressor <class 'sklearn.linear_model._glm.glm.GammaRegressor'>
GaussianProcessRegressor <class 'sklearn.gaussian_process._gpr.GaussianProcessRegressor'>
...

How can one then import the class object within the <>'s? So I end up with something like (psuedo):

# Print/import all regressors
estimators = all_estimators(type_filter="regressor")
for name in estimators:
    print(name[0], name[1])
    import name[1]

...but of course that doesn't work. Thanks!

UPDATE What I'm wanting to do is then use the imported class. Trying something like:

for name in estimators:
    globals()[name[0]] = name[1]
    params = name[0].get_params()
    print(params)

... doesn't work.

CodePudding user response：

You've got a name and a class, so you could just add them to your namespace

for name in estimators:
    globals()[name[0]] = name[1]

But a dictionary may be better

my_estimators = {name[0]:name[1] for name in estimators}

I don't know how these classes are instantiated, but suppose they take two paramters and you know the name of the one you want. Then you would do

my_estimator = my_estimators["ARDRegression"]("foo", "bar")
my_estimator.get_params()

CodePudding user response：

You can simply turn the return value from all_estimators into a dictionary whose keys are the estimator names, and the values are the estimator class themselves, which you can directly instantiate:

from sklearn.utils import all_estimators

estimator_dict = dict(all_estimators(type_filter="regressor"))

# we get and instantiate the an ARDRegressor in the line below
ard_regression_estimator = estimator_dict['ARDRegression']()

print(ard_regression_estimator)  # output: ARDRegression()