[FIXED] How to dynamically instantiate sklearn classifiers by providing the name as a string

Issue

I would like to instantiate a sklearn estimator thanks to its name. For example :

name = 'RandomForestClassifier'
clf = BaseEstimator(name)
print(clf) # return something like RandomForestClassifier(...)

What I have tried is clf = eval(name) but I can't do clf.fit(X, y) because clf is <class 'sklearn.ensemble._forest.RandomForestClassifier'> instead of something like RandomForestClassifier().

I can't find how to create a sklearn estimator thanks to its name.

Solution

As far as I understand, what you want to achieve is to dynamically instantiate existing classifiers based on some user input given in string. BaseEstimator is the base class for all estimators so I think you should use it only in case you are planning to build your own Estimator.

---

Now I don't know how did you manage to run the above code as BaseEstimator() doesn't take any arguments:

clf = BaseEstimator('RandomForestClassifier')
Traceback (most recent call last):
  File "/usr/local/lib/python3.7/site-packages/IPython/core/interactiveshell.py", line 3331, in run_code
    exec(code_obj, self.user_global_ns, self.user_ns)
  File "<ipython-input-6-59dd1a62618a>", line 1, in <module>
    clf = BaseEstimator('RandomForestClassifier')
TypeError: BaseEstimator() takes no arguments

---

In any case, the closer you can get to a dynamic instantiation of classifiers is the following:

from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier


def get_clf(name):
    if name not in ('RandomForestClassifier', 'DecisionTreeClassifier'):
        raise ValueError(f"{name} is not a recognised option")    

    classifiers = {
        "DecisionTreeClassifier": DecisionTreeClassifier,
        "RandomForestClassifier": RandomForestClassifier
    }

    classifier = classifiers[name]
    return classifier()

Alternatively, you can go with a more straight forward solution:

def get_clf(name):
    
    if name not in ('RandomForestClassifier', 'DecisionTreeClassifier'):
        raise ValueError(f"{name} is not a recognised option")

    if name == 'RandomForestClassifier':
        return RandomForestClassifier()
    elif name == 'DecisionTreeClassifier':
        return DecisionTreeClassifier()
    else: 
        raise RuntimeError()

Answered By - Giorgos Myrianthous