making Class inheritance as an argument

user40780 :

Consider this simplest possible case where we want our class to inherit from any sklearn clustering class:

class MyOwnClustering(AgglomerativeClustering):
    def __init__(self,mysetting , **kwargs):
        super().__init__(**kwargs)

Is it possible to make AgglomerativeClustering as an argument? So that I can switch freely from any clustering algo (like Kmeans)?

Edit:

Based on 53RT's method: can i do this instead?

def construct_clustering(cls):

    class MyOwnClustering(cls):
        def __init__(self, mysetting , **kwargs):
            super().__init__(**kwargs)

    return MyOwnClustering

Tested, yes I can.

53RT :

I think your problem can be solved using the Builder pattern which is showcased here.

This would yield the following solution.

from sklearn.cluster import AgglomerativeClustering, KMeans
from sklearn import datasets

setting = dict()
noisy_moons = datasets.make_moons(n_samples=50, noise=.05)

You wrap your creation into a function

def construct_clustering(cls, setting):

    class MyOwnClustering(cls):
        def __init__(self, mysetting , **kwargs):
            super().__init__(**kwargs)

    clustering = MyOwnClustering(setting)

    return clustering

And can create instances of MyOwnClustering with different inheritance

m = construct_clustering(AgglomerativeClustering, setting)
print(isinstance(m, AgglomerativeClustering))

c = m.fit(noisy_moons[0])
c.labels_

>>> array([0, 0, 0, 1, 1, 0, 0, 1, 1, 0, 1, 1, 0, 0, 0, 0, 1, 0, 0, 0, 1, 1,
       1, 0, 0, 1, 0, 0, 0, 0, 1, 1, 1, 1, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0,
       0, 1, 1, 0, 0, 1], dtype=int64)

m = construct_clustering(KMeans, setting)
print(isinstance(m, KMeans))  # True

c = m.fit(noisy_moons[0])
c.labels_

>>> array([6, 6, 2, 1, 5, 0, 0, 3, 7, 0, 5, 1, 6, 4, 4, 4, 1, 0, 2, 0, 7, 5,
       7, 0, 4, 1, 4, 6, 0, 2, 7, 5, 3, 7, 5, 2, 6, 6, 5, 2, 6, 7, 4, 2,
       4, 3, 3, 3, 0, 1])