Dissertations, Theses, and Capstone Projects

Date of Degree

2-2025

Document Type

Dissertation

Degree Name

Ph.D.

Program

Physics

Advisor

Robert Haralick

Committee Members

Alexandre Bovet

Miguel Fiolhais

Olympia Hadjiliadis

Hernan Makse

Visvanathan Ramesh

Subject Categories

Physics

Keywords

complex systems, network science

Abstract

The first chapter introduces the fundamentals of network science. After defining the necessary terminology and background material for this discussion, we detail several important types of networks, network centrality metrics, and processes for which these centrality metrics are useful. The second chapter derives a directed variant of the Collective Influence (CI) algorithm. We demonstrate its efficacy in breaking the strongly-connected component of directed networks compared to other centrality metrics. We then revisit the original CI algorithm, correcting an error in its derivation to dramatically improve its effectiveness. Finally, we provide an optimized method for computing CI. The final chapter introduces a toy model of an economy and attempts to use our directed Collective Influence (dCI) algorithm as a dimensional reduction method for predicting the dynamics of this economy. To validate this approach against actual data, we consider the Bitcoin transaction network. Using both Transfer Entropy and the Correlation Ratio to try to predict the price of Bitcoin on three time periods based on the activity of the three principal influencers identified by dCI in each time period, we fail to discover any predictive relationships. Alternative approaches are discussed, but not explored.

Download

This work is embargoed and will be available for download on Monday, February 01, 2027

Included in

Physics Commons

Share

COinS