I would like to perform a clustering (in the best case scenario a hierarchical clustering) of N entities and the distance among those entities is a known input. I also have an a priori on the relationship entity-cluster and the total number of clusters. Those two are not coming from the distance matrix I start from, for the sake of simplicity let's say it is a completely exogenous piece of information. Here my questions:

  • hierarchical clustering typically provides as an output the complete dendogram without providing any specific information on the best cut. Are there hierarchical clustering methods which allow to specify a prior for the best cut (i.e. my a priori entity-cluster) or in general a hierarchical clustering method for which it would be possible to specify some a priori information.
  • if no hierarchical clustering methods fit this picture, which clustering would you recommend for my specific use case?
  • cherry on the cake: an off-the-shelf (or close to) solution in python to explore this clustering + prior, it is a project in a very early exploration phase, the clustering at this stage it is more a tool to have an insight than the objective of the exploration.
  • 1
    $\begingroup$ Clustering is a method without using a prior information about which cluster which point belongs to. Because there are no clusters yet, prior clustering. One could impose a prior constraint regarding whether any two points may or may not (or should or should not) be clusteted together, - but this is a constraint for the distance matrix. $\endgroup$ – ttnphns Feb 5 at 21:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.