PhD Dissertation Defense by: Shaon Bhatta Shuvo

 

Complex systems, characterized by their intricate, interconnected components and emergent behaviours, pose significant challenges for analysis and optimization. These systems are prevalent across various domains, from biological ecosystems to social networks and technological infrastructures, requiring sophisticated computational approaches to understand and manage their dynamics effectively. Addressing this challenge, this dissertation harnesses Artificial Intelligence (AI) methodologies to analyze and optimize complex systems, with a focused application in computational epidemiology. This domain, exemplifying complex system dynamics through the spread of infectious diseases, provides a rich context for deploying a range of AI techniques, including agent-based modelling, mathematical modelling, machine learning, and deep reinforcement learning.

 

The broad objective is twofold: first, to enhance predictive and analytical capabilities in computational epidemiology, thereby refining decision support systems within public health; and second, to advance the development and application of optimization algorithms and modelling frameworks, within computer science. By employing a diverse set of computational techniques, this research achieves a nuanced understanding of the interactions and patterns governing infectious disease dynamics. This comprehensive approach not only informs the development of more effective public health strategies and interventions but also demonstrates the potential of computational innovations to guide real-world applications. These case studies serve as a practical demonstration of how AI methodologies can be applied to real-world challenges, showcasing the direct impact of computational advances on complex decision-making.

This dissertation lays the groundwork by simulating the implications of healthcare capacity and social distancing measures, illustrating their potential to mitigate the spread of the virus. Progressing further, the research digs into predictive modelling, utilizing advanced artificial agent-based simulation techniques to forecast the trajectory of the pandemic, including active cases and hospitalizations, and further offers a granular analysis of COVID-19's impact across Canada through deep learning models by building a new multi-factored dataset. The dissertation takes a significant leap forward by integrating compartmental models, agent-based modelling, and deep reinforcement learning to formulate a decision support system and assess optimal intervention strategies. This effort addresses a significant optimization challenge, marking a critical advancement in this domain. In essence, this innovative blending of methodologies not only advances the field of computational epidemiology but also advances the core objectives of computer science, driving technological innovation that yields tangible societal benefits.