DataPathway
The problem
Network congestion and wasted bandwidth come from inefficient paths between devices. For the Algorithmic Complexity course, we modeled a real dataset of over 3.5 million network flows as a graph to find the cheapest set of connections linking its devices.
Key decisions
- Modeled IP addresses as nodes and the traffic between them as weighted edges.
- Implemented both Kruskal and Prim to build the minimum spanning tree, since their efficiency depends on how sparse or dense the network is.
- Measured algorithm and drawing time separately for each run, so rendering cost wouldn't distort the comparison.
- Wrapped the analysis in a Flask web page that compares both algorithms and draws the resulting graphs.
The result
A tool that builds both spanning trees from the flow data and shows, side by side, the edges kept and the time each algorithm took.