Despite its popularity, the Elbow method is pretty much the worst choice one can do when setting the number of clusters for a dataset

https://link.springer.com/chapter/10.1007/978-3-540-88458-3_60#:~:text=Bayesian%20Information%20Criterion%20(BIC)%20is,as%20the%20number%20of%20clusters.

For Sklearn app:

https://stats.stackexchange.com/a/251169/165101