Graph Algorithm Visualization Tool

Our idea is to visualize the process of how community detection algorithm capture the graph structure and compare the performance of different type of community detection algorithm

Comment 1

Inspiration

Recently we are doing some projects in the field of auto-insurance fraud detection. The behavior of insurance fraud is usually performed by a cluster of people. As we all know, community detection is one of the most important kind of graph algorithms that can be applied in this scenario. To get a full understanding of how community detection algorithm works and performance comparison between different types community detection algorithms, we develop this visualization tool.

What it does

With our visualizaton tool, you can upload your own csv files to define a graph structure. By clicking the submit button, you can see the each-step results of Girvan-Newman algorithm and modularity variation. Furthermore, the final results of other kinds of community detection algorithm such as Louvain are also illustrated below.

How we built it

Our project consists of two main parts: 1) back-end computation unit 2) front-end visualization UI interface. The first part is built with networkx python library and the second part is built with angular and d3 library.

Challenges we ran into

The main challenge that we have encountered is to find a proper way to demonstrate the computation process of community detection.

What we learned

We have become proficient with networkx, angular and d3 libraries and learned that collaboration is the key element for a team to overcome difficulties.

What's next for Community Detection Visualization Tool

Next we will extend our visualization tool to include more types of graph algorithms such as PageRank and add some functionality to interact with Neo4j in a better way.

+ 2 more
Share this project:

Updates