When
Thursday, December 5, 2024 - 2:00 p.m.
Fenglian Pan
PhD Candidate
Systems and Industrial Engineering
University of Arizona
"Analyzing Discrete Spatio-Temporal Event Sequence for Knowledge Discovery"
ENGR 301
Abstract: Discrete events are a sequence of observations consisting of event time, location and possibly contextual attributes with additional event information. Such data is prevalent in various areas, including failure events in complex systems, disease outbreaks in healthcare and traffic demands in transportation networks. Accurate prediction of these events is crucial for effective system management, yet it is hindered by three main challenges: 1) intricate interdependencies among events, 2) a wide range of contextual attributes that influence the event occurrence, and 3) the scalability required to analyze large-scale datasets. In this seminar, I will present our recent research addressing these challenges based on stochastic point processes. These approaches explicitly model the effects of event interdependencies and contextual attributes, enabling a comprehensive understanding of event dynamics. Additionally, I will introduce a computationally efficient model estimation procedure with statistical guarantees, designed to handle large-scale data analysis. Multiple case studies were conducted to demonstrate the effectiveness of the proposed methods, including AI systems reliability modeling and opioid overdose prediction in healthcare.
Bio: Fenglian Pan is a PhD candidate in systems and industrial engineering at the University of Arizona, advised by Professor Jian Liu. Her research centers on discrete spatio-temporal (ST) event data analytics, with a focus on developing advanced methodologies in statistical modeling, causal inference and simulation. Through her work, she aims to improve the accuracy of ST event occurrence predictions and thus provide actionable insights for better system management. Her research has been applied in multiple areas, including reliability modeling for Artificial Intelligence systems, opioid overdose prediction for health informatics and traffic demand prediction in transportation networks.