Examining the Predictive Power of Clustering Coefficients on Various graphs With Complex Fractional Contagion

Abstract

Complex group behavior is an important but little understood area of research with applications both in biology as well as technology. Modeling these groups as graphs leads to an easier way to understand behavior but potentially overly simplifies actual natural behavior. Using golden shiner schools as an example of a social group, I used simulations to examine how various graph models affect the behavior of information flow through these networks. I find that only weighted undirected graphs are able to replicate the information propagation behavior observed in nature. Furthermore, I find that the many low weighted edges found in the natural golden shiner networks are crucial for the type of information propagation observed. Even thresholding out edges that in total represent just 0.5 percent of the total edge weighting in a graph creates a large change in information propagation behavior.