Decision tree and clustering?

Hi,
In quiz M5.01 you say that decision trees can t not be used for clustering tasks but visibly not everyone is agree with you.
In a quiz on another site about datascience :

or on the WEB site of the UCL :

Decision trees are mainly used to perform classification tasks. Samples are submitted to a test in each node of the tree and guided through the tree based on the result. … Decision trees can be extended to clustering problems with the constraint that labels of the samples are unknown during the training process.

And I even found an scientific article called “Clustering Through Decision Tree Construction

Can you comment?

I’m finding more and more exemples of use of decision trees fro clustering tasks. The most current use is to use decision tree to do the clustering. Using decision trees in the making of the clustering help a lot to the interpretation.
So I’m really wondering why you say that

they cannot be used for unsupervised tasks such as clustering

.??

We wanted to have a trivial question and we failed :slight_smile:

Indeed, decision trees could be used for some unsupervised learning as for instance the RandomTreesEmbedding

We intended to have a simple question following the video where we only discuss supervised learning aspects. We will probably remove and improve add a new question in the next version of the MOOC.

Perhaps you could modified the question as:

For which kind of supervised tasks Decision trees can be applied :

Like that you have still a simple question and you have just to say in the solution that clustering are not supervised and c is not a good answer :grinning_face_with_smiling_eyes:

2 Likes

I made a rephrasing here: FIX make question less ambiguous · INRIA/scikit-learn-mooc@cb1844c · GitHub

The question will be updated in FUN soon.

@lfarhi @MarieCollin Could you update FUN with the following change: https://gitlab.inria.fr/learninglab/mooc-scikit-learn/mooc-scikit-learn-coordination/-/commit/e5783eaec576d6d8edb02f947cecad3c42275be5

It’s fixed also in FUN

:+1: :+1: