[ENH] Implementing DBSCAN

[aryanpola]

Describe the feature or idea you want to propose

We currently do not have a sub-module for density-based clustering. This issue is for implementing Density-Based Spatial Clustering of Applications with Noise (DBSCAN).

Describe your proposed solution

Paper Reference: https://cdn.aaai.org/KDD/1996/KDD96-037.pdf

Scikit-learn for reference: https://github.com/scikit-learn/scikit-learn/blob/main/sklearn/cluster/_dbscan.py

Describe alternatives you've considered, if relevant

No response

Additional context

The scikit-learn implementation cannot predict. This does not fit with our interface.

[aryanpola]

@aeon-actions-bot assign @aryanpola

[SebastianSchmidl]

DBSCAN is already implemented in sklearn, and aeon depends on sklearn. Why do we need to re-implement it directly in aeon?

[aryanpola]

The clustering examples had a graph contained all the clustering sub-modules which includes density based, so figured why not implement it in aeon.
Please let me know if this is unnecessary, I'll close this.
Thanks for checking!

[MatthewMiddlehurst]

Agreed that unless there is some reason the sklearn version wont work (or is insufficient in any other way) that theres no need to have it here really. Any thoughts @chrisholder ?

[TonyBagnall]

the problem with dbscan is that it does not have a predict, i.e. it can only cluster train data. I'm not sure of the best way to deal with this. I'll have a go with the scikit learn version

[aryanpola]

the problem with dbscan is that it does not have a predict, i.e. it can only cluster train data. I'm not sure of the best way to deal with this. I'll have a go with the scikit learn version

Yep, not having a predict function makes it a lot more difficult to cluster the newer points into the existing clusters (requires reevaluation of the model).
I'll close the issue then.