I am looking for resources on tuning HDBSCAN for vehicle routing problems for multiple depots.
In this case, the points being clustered are locations with latitude and longitude. I'd like to identify at least K clusters (being the number or depots). While HDBSCAN seems not to be able to provide the K clusters, I can post-process to split and merge clusters.
From the documentation, I have started playing around with the 3 parameters - min_cluster_size, min_samples and cluster_selection_epsilon.
Hoping for advice on how to set the parameters to get a set of clusters for the routing algorithm to work. The ideal set of clusters would allow for cost optimal routes to be created.
