Twitter Community Detection

Finds Communities within the Twitter domain using Spectral Clustering and Influence.

Steps:

1. Harvest Tweets using streaming API (stream.py), store tweets in MongoDB
2. Remove any tweets that do not have a retweet/mention
3. Construct a graph from the remaining tweets
4. Use Spectral Clustering to find the best clusters
5. Evaluation Clustering using the Silhouette score

Requires the following modules:

• tweepy
• pymongo
• networkx
• matplotlib
• numpy
• sklearn

Peter J. Rousseeuw (1987). “Silhouettes: a Graphical Aid to the Interpretation and Validation of Cluster Analysis”. Computational and Applied Mathematics 20: 53–65. doi:10.1016/0377-0427(87)90125-7.