Online social networks have gained tremendous attention. People learn new knowledge from their online role models, and reshare information causing cascades of online information sharing. Disinformation can also be shared rapidly, and it is difficult to disambiguate the real the fake news on OSNs. Traditional social behavioral theories often fail to fully explain social behavior online due to the discrepancy between how people communicate online versus offline. Modeling information exchange and propagation on OSNs is critical across a variety of domains from business to politics. Many approaches to modeling online social behavior leverage manual pattern matching, semantic networks, and traditional machine learning techniques, where the estimated modeling itself is static. This dissertation proposes a temporal perspective to examine the patterns of online social behavior with deep neural network learning based approaches. The objective of this dissertation is to implement a deep network learning framework that effectively addresses the temporal aspect of online social behavior. The dissertation consists of three articles. All of these articles study online social behavior in a specific context and each one focuses on a different aspect of the online social behavior. Chapter 4 tests the ability of recurrent neural networks to detect online disinformation in financial text data. This study used a temporal recurrent neural network to simultaneously model textual and temporal features and examine their relationships with stock price movement to gain a deeper understanding of how disinformation effects online social behavior. Chapter 5 examined the impact of "influencer effects" in distributed project management. Based on social learning theory, this study utilized deep network dynamics to examine how people learn from their role models in the form of triadic effect. Chapter 6 considers the diffusion aspect in online social behavior and proposes a novel temporal cascade deep network learning model to identify the depth, breath and scale of the diffusion process. In the proposed model, large-scale high-fidelity cascades are simulated to illustrate these sophisticated interactions within different populations. This overarching goal of this dissertation is to model the following: online social behavior in a variety of domains, the effects of influencers on information dissemination, and to quantify the capability of disinformation detection via state-of-the-art recurrent neural networks.


If this is your thesis or dissertation, and want to learn how to access it or for more information about readership statistics, contact us at STARS@ucf.edu

Graduation Date





Li, Yao


Doctor of Philosophy (Ph.D.)


College of Engineering and Computer Science


School of Modeling, Simulation, and Training

Degree Program

Modeling & Simulation


CFE0009737; DP0027845





Release Date

August 2028

Length of Campus-only Access

5 years

Access Status

Doctoral Dissertation (Campus-only Access)

Restricted to the UCF community until August 2028; it will then be open access.