I shall be thankful if someone guide me and also share some related study material.I need your help and guidance for my PhD thesis topic. For my PhD thesis, I am looking for a thesis topic and proposal that will perform community detection on the Telegram groups and channels that I identified. Data that I have available : More than 2 million groups More than 6 million channels More than 300 million users who are members of these groups More than 2 billion records related to individual membership. Almost every person is a member of 7 groups on average This is basic information What method or approach do you suggest I consider for my PhD proposal and what do you think the research questions could be?
Dear Fateme Dsjd
The dataset you have is really impressive and rich, and it opens up many exciting research possibilities in graph theory and network science beyond community detection.
Here are some research directions in graph theory you might consider for your PhD:
1. Influence Maximization and Information Diffusion
Explore how information spreads across Telegram groups/channels. Who are the most influential users? What patterns can be observed in viral content spread?
2. Link Prediction and User Recommendation
Based on group/channel membership, can we predict future group joins or suggest relevant communities to users?
3. Temporal Network Analysis
Since you likely have time-stamped data, consider studying how the network evolves over time. For instance:
- How do groups grow or shrink?
- What are the dynamics of user engagement?
🔹 4. Anomaly or Bot Detection
With such a large user base, detecting fake users or bot-like behavior using graph-based anomaly detection would be a meaningful and practical contribution.
5. Hierarchical or Overlapping Community Structures
If you're still interested in community detection, you could go beyond traditional approaches and explore overlapping or hierarchical communities, which often reflect real-world social structures better.
6. Hypergraphs or Multilayer Networks
Traditional graphs might be limiting. Consider modeling your data as a hypergraph (users belong to multiple groups at once) or as a multiplex network (e.g., user-group, user-channel, user-user layers).
Possible Research Questions
- How do structural properties of Telegram's network influence information flow?
- Can user behavior patterns predict the formation or dissolution of groups?
- Are there identifiable clusters of influence or control across the Telegram network?
- How do temporal and structural aspects interact in community lifecycles?
I hope the content I wrote proves to be useful.
There are two papers that came to my mind. First is 'Graph structure in the Web' by Broder, Kumar, Maghoul, Raghavan, Rajagopalan, Stata, Tomkins and Wiener. The second is 'The Maximum degree & diameter-bounded subgraph and its applications' by Dekker, Perez-Roses, Pineda-Villavicencio and Watters. Analysing the Telegram network along these lines would be interesting. I wonder also how the structure of the network changed after Telegram removed the discover people nearby feature?
One point that I would make though is that in most countries a maths PhD project proposal is made by the supervisors, as it is difficult to know how to come up with a good question when you are just starting your studies. Maybe it works differently in Iran (if that is where you are studying)?
By the way, does the tag need to be 'community' rather than 'communism'? :D
- Badges
- Science method