A rooted-pagerank based link recommendation system for directed graphs. Specific app which scrapes github and recommends people to follow.
- Flask-based web server, and Bootstrap as frontend.
- Uses Personalised or Rooted Pagerank to find rank prospective *folowees*
- The rank of kth node is obtained based on the kth element of the eigen probability vector for the random walk with teleport to root matrix.
- networkx used for graph operations, and numpy for solving for the eigenvalues.
Needs python3 and venv submodule. Run setup.sh
Run start.sh. Visit 127.0.0.1:5000
-
Scraping data from github takes time for slow connections. So program caches user's networks in the data/ folder. If such a file exists, it does not scrape data. This means that the recommendations will not change even if changes occur in the github network. Delete the data file for updation.
-
The app is NOT secure, and does not handle all possible paths well. Misuse is not supported.