Author ORCID Identifier
https://orcid.org/0009-0005-9930-9204
Defense Date
2026
Document Type
Dissertation
Degree Name
Doctor of Philosophy
Department
Systems Modeling and Analysis
First Advisor
Yanjun Qian
Second Advisor
Michael Robert
Third Advisor
James Mays
Fourth Advisor
Derek Johnson
Abstract
Dengue is a mosquito-borne disease that poses a public health threat to over half the world’s population, affecting an estimated 100 to 400 million people across 129 countries. Although dengue is prevalent in tropical and subtropical regions, cases have emerged in temperate areas recently, driven in part by climate change and globalization. Developing accurate and timely predictive models for dengue is important to improving outbreak preparedness and response. In this dissertation, we explore novel, data-driven modeling approaches to understand and predict dengue spread in different environments.
In the first project, we explore the relationship between Google search terms and observed dengue cases in Córdoba, Argentina, a province where dengue first emerged in 2009. Specifically, we identify the important search terms that affect dengue cases through various statistical and machine learning methods. We consider real-time and lag scenarios and compare the accuracy of all methods to determine the model that performs the best in terms of model fitting and prediction. In the second project, we extend our first project using Bayesian analysis. Specifically, we consider a two-part hybrid model: first, we determine whether an outbreak would occur based on Google search terms and climate data; second, we identify the important Google search terms and climate data conditioned on whether an outbreak occurs. In the third project, we focus on two recent dengue outbreaks in the Dominican Republic, where dengue has been endemic for over thirty years. We propose a neural network architecture to model dengue cases as a function of time, previous cases, and climate. We employ artificial neural networks (ANNs) and physics-informed neural networks (PINNs) to study dengue transmission in the key province of Barahona in the country. In both the ANN and PINN frameworks, we explore one- and two-week prediction horizon scenarios to compare weekly vs. biweekly predictions.
Taken together, the work in this dissertation contributes novel approaches to advancing computational, mathematical, and statistical models for better understanding dengue spread in emerging and endemic regions. Within each chapter, we discuss the implications of our work for dengue forecasting and the development of Early Warning Systems that can be used to improve the prediction of the timing and magnitude of outbreaks. This work is useful for informing public health and mosquito control agencies, and although this dissertation focuses on dengue, the modeling approaches are adaptable to other diseases.
Rights
© The Author
Is Part Of
VCU University Archives
Is Part Of
VCU Theses and Dissertations
Date of Submission
5-8-2026
Included in
Applied Statistics Commons, Data Science Commons, Epidemiology Commons, Infectious Disease Commons, Statistical Models Commons