Date, Time, Location: July 6, 2023, 10 am, COB 320 (the Willow Conference room)
Title: Machine Learning and Data Science for System Identification, Public Health, and Equity.
Abstract: Machine learning and data science are extremely powerful tools for prediction, analysis, system identification, and estimation. This dissertation presents two new methods for system identification using neural networks. The first method, covered in Chapter 2, uses neural shape functions to learn vector fields from data. The second, presented in Chapter 5, integrates neural networks into a Continuous Time Markov Chain framework to learn propensity functions from data. This integration allows for propensity functions which depend on outside covariates. In addition to proposing these methods, this dissertation also examines fairness in machine learning algorithms and proposes preprocessing methods to increase fairness and equity in machine learning algorithm outputs. These methods, Separate and Equity resampling, are introduced in Chapter 3 in the context of predicting suicide death. The Equity resampling method is further examined in Chapter 4. The overarching goal of this dissertation is to present new methodologies in machine learning and data science to solve problems using real world data.
