Skip to content

Latest commit

 

History

History
23 lines (15 loc) · 1.48 KB

README.md

File metadata and controls

23 lines (15 loc) · 1.48 KB

parallelized-dc

A Parallel Graph Partitioning Approach to Enhance Community Detection in Social Networks

The need to detect communities in social networks has always been challenging, especially when analyzing large networks. Therefore, I present a new partitioning algorithm based on what I call a subtree-splitting strategy. The algorithm was designed to work on density-based algorithms such as NetSCAN or DBSCAN. The algorithms' goal is to split a graph structure into n smaller components with respect to the following particulars:

  • The partitioning process keeps the partition's connectivity.
  • The algorithm strategy also allows for the overlapping of nodes between partitions.

Published Work

This work was published in the 25th IEEE Symposium on Computers and Communications (ISCC).

Database

The Digital Bibliography& Library Project (DBLP) database was modeled as a scientific citation network used to support the analysis and experiments.

Open Challenges

The partitioning method is working and well commented. However, since the main goal is to speed up the clustering process, we've been working on the parallelism process followed by the merge implementation.