HNDS-R: Extracting the Backbone of Unweighted Networks

Michigan State University, East Lansing MI

Investigators

Abstract

In this project, methods are developed for extracting the backbone from dense, unweighted networks to facilitate their analysis. Networks influence all domains of society, including the diffusion of information and misinformation, the contagion of illness, the passage of legislation, the emergence and maintenance of norms, and the formation of close relationships. In many cases, these networks can be difficult to analyze because they contain so many relationships, and because the strength of these relationships is unknown. Backbone extraction involves identifying and retaining only the most important relationships, which yields a simpler network that can be more readily analyzed and visualized. In this project, in addition, the widely employed backbone software is extended to include these newly developed methods so that researchers can use them easily. This work facilitates the analysis of networks that arise in many different contexts and are studied in many different fields. It also involves the development of training materials to guide researchers in the selection of backbone methods. Development of methods for extracting the backbone from unweighted networks proceeds in three stages. First, the common steps involved in existing backbone extraction methods are identified. Next, each of these steps is implemented into a new function in the backbone package for R. This function allows for the application of both existing backbone models and new backbone models described by novel recombinations of common steps. Finally, the performance of each existing and each new backbone model is evaluated by extracting the backbone from simulated dense unweighted networks that have been embedded with hidden community or hub structures. Once the most promising backbone models are identified, their software implementations are refined for scalability, allowing them to be applied to large networks. Additionally, software documentation and training materials are prepared that provide guidance to researchers using backbone models and the associated backbone package. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.