August 1, 2017.
Community detection is the process of analyzing graphs to distinguish different groups of nodes from one another. A community is defined as a group of nodes that are closely related among each other but loosely related to the other nodes in the network. These communities exist within the species, gene, and protein networks of a microbiome. Many different algorithms have been developed to detect these communities. The project as a whole is intended to track communities in dynamic networks using known community detection algorithms. An initial effort created implementations of different algorithms for community detection to test for community quality with respect to computational time, focusing on the Girvan-Newman algorithm and the Louvain algorithm. Trials were run on assortative planted partition models to test the accuracy of the algorithms with respect to their computational time. After the trials, the Louvain algorithm was identified to not only be more computationally-time efficient, but more accurate when detecting communities in models with less assortativity. The accuracy and efficiency of the Louvain algorithm is promising for its future use in dynamic community detection in networks that model microbiomes in transition. Preliminary detection efforts on dynamic networks with community structure were performed on models using the framework of the Chinese Restaurant stochastic process. These efforts attempted to track community structure over time, utilizing the Jaccard index and Pointwise Mutual Information, or PMI. Leveraging these preliminary results, we plan on developing a set of formal rules to track communities in dynamic graphs.
Pacific Northwest National Laboratory (PNNL)
This material is based upon work supported by the National Science Foundation through the Robert Noyce Teacher Scholarship Program under Grant #1418852. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation. The research was also made possible by the California State University STEM Teacher and Researcher Program, in partnership with Chevron (www.chevron.com), the National Marine Sanctuary Foundation (www.marinesanctuary.org), and Pacific Northwest National Laboratory.