Solving the Inverse Problem Using Combination Random Graph Models
Abstract
We seek to determine if real-networks can accurately be represented by random graph models. To accomplish this task, we use a combination of three commonly-used random graph models: geometric, Chung-Lu, and preferential attachment. Each of these three models has unique properties that helps model certain characteristics of real-world networks, but using these random graph models individually has proven fruitless. Therefore, we combine multiple models in order to get a model that more accurately reflects these networks. Our method for determining if our combination random graph model successfully represents a real-world network consists of three main tests: edge counts, degree distributions, and triangle counts. This developed algorithm supports the idea that random graph models have potential in modeling real-world networks, and its output is further supported by statistical tests we develop. Although we find some faults in our method, it shows significant potential. We achieved some success with organically produced real-world networks like human and animal social networks and terrorist cells. However, we hypothesize that our model can be improved by adding more random graph models and testing it on larger networks.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 20, 2019
- Accession Number
- AD1073938
Entities
People
- Samuel H. Baker
Organizations
- United States Naval Academy