Theses and Dissertations
Date of Award
5-1-2026
Document Type
Thesis
Degree Name
Master of Science (MS)
Department
Mathematics
First Advisor
Kristina Vatcheva
Second Advisor
Zhuanzhuan Ma
Third Advisor
Thoa Thieu
Abstract
Obesity affects over 40% of U.S. adults and is associated with substantial morbidity. Metabolic and bariatric procedures, including Roux-en-Y gastric bypass and sleeve gastrectomy, produce significant and durable weight loss, yet individual postoperative trajectories vary considerably. Accurate prediction of weight outcomes is therefore critical for patient counseling and clinical decision-making. Longitudinal bariatric datasets are typically unbalanced and time-unstructured, posing challenges for conventional modeling approaches. While statistical methods such as linear mixed-effects models, generalized estimating equations, generalized additive mixed models, and RE-EM trees are widely used for correlated data, machine learning and deep learning techniques offer flexible alternatives capable of modeling complex nonlinear and temporal patterns. In this study, we compare traditional statistical approaches with mixed-effects random forests and neural network architectures (ANN, MLP, RNN, LSTM, GRU) to predict postoperative weight trajectories. Our findings highlight performance trade-offs between statistical and ML/DL models in irregular longitudinal settings.
Recommended Citation
Gutierrez, D. (2026). Predictive Statistical and Machine Learning Approaches for Time-Unstructured and Unbalanced Longitudinal Data: Application to Bariatric Surgery [Master's thesis, The University of Texas Rio Grande Valley]. ScholarWorks @ UTRGV. https://scholarworks.utrgv.edu/etd/2001

Comments
Copyright 2026 Daniel Gutierrez III. All Rights Reserved.