Detecting Communities in Social Graph by Ricky Ho.
From the post:
In analyzing social network, one common problem is how to detecting communities, such as groups of people who knows or interacting frequently with each other. Community is a subgraph of a graph where the connectivity are unusually dense.
In this blog, I will enumerate some common algorithms on finding communities.
First of all, community detection can be think of graph partitioning problem. In this case, a single node will belong to no more than one community. In other words, community does not overlap with each other.
When you read:
community detection can be think of graph partitioning problem. In this case, a single node will belong to no more than one community.
What does that remind you of?
Does it stand to reason that representatives of the same subject, some with more, some with less information about a subject, would exhibit the same “connectivity” that Ricky calls “unusually dense?”
The TMDM defines a basis for “unusually dense” connectivity but what if we are exploring other representatives of subjects? And trying to detect likely representatives of the same subject?
How would you use graph partitioning to explore such representative?
That could make a fairly interesting research project for anyone wanting to merge diverse intelligence about some subject or person together.