How to measure the distance between two data points (or: nodes?) based on their mutual share of linkages?

I don't know the technical term for that, so here is a fictitious example from scientific publishing:

  • Say that 20% of all citations from the journal Acta Mathematica go to the journal Nature.

  • Only 5% of all citations from Acta Mathematica go to the Romanian Journal of Physics.

  • On the other hand, Nature never cites Acta Mathematica, ...

  • ... while 10% of the Romanian Journal of Physics' citations go to Acta Mathematica.

citing cited share
Acta Mathematica Nature 15%
Acta Mathematica Romanian Journal of Physics 8%
Nature Acta Mathematica 0%
Romanian Journal of Physics Acta Mathematica 10%

Given such inter-linkages, the Romanian Journal of Physics and Acta Mathematica are "closer" to each other than the Romanian Journal of Physics and Nature.

Is there a conventional way to calculate such distances?

I believe it would be too simplistic and wrong to just sum the mutual shares? I mean in the form of:

journal1 journal2 sum(share)
Acta Mathematica Romanian Journal of Physics 18%
Acta Mathematica Nature 15%

This seems too un-sophisticated and counterintuitive. If Acta Mathematica's citations to Nature would amount to 20%, then Nature would be "closer" to Acta Mathematica than the Romanian Journal of Physics is? This doesn't sound right to me.

I would be grateful for any hints regarding algorithms that would "objectively" measure such distances!

  • $\begingroup$ This is definitely a graph problem. There are certainly many methods for calculating distances in a weighted graph, as well as clustering methods to group nodes together. There might also be a probabilistic angle to it. $\endgroup$
    – Erwan
    Commented Nov 27, 2021 at 16:29


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.