I'm looking for the most correct way to calculate a proximity score between users of a social network (Twitter). There are 100 users who replied, mentioned, and liked each other. I want to aggregate these measures into a proximity score between each pair of users. Sounds like a typical dimension reduction task. Basically, I'm trying to elicit the measure of how close the relations between the two people are.
Currently, I'm using PCA. Are you aware of any SNA publications that did something like this? Or how would you do this?
Would you maybe weigh individual proximity scores with the overall network activity of the person? For example, user A only follows user B and liked her posts 5 times, while user C follows 100 users and also liked B's posts 5 times. Then, A should score closer to B compared to C.
A link to such a research will be much appreciated.
