There are iterative methods for optimizing the number of clusters in k-means (silhouette and knee/elbow are common), but in practice I prefer density-based meth

There are iterative methods for optimizing the number of clusters in k-means (silhouette and knee/elbow are common), but in practice I prefer density-based methods like HDBSCAN and OPTICS. There's a very basic visual comparison at https://scikit-learn.org/stable/auto_examples/cluster/plot_cluster_comparison.html#sphx-glr-auto-examples-cluster-plot-cluster-comparison-py https://scikit-learn.org/stable/auto_examples/cluster/plot_c....