Preliminary Findings: Leveraging Machine Learning Models to Analyze Statewide Chronic Disease Measures that Predict Geographic Regions Across U.S. States
Poon, Livi
Poon, Livi
Citations
Altmetric:
License
License
Abstract
Chronic disease is a leading cause of death in America, with heart disease, cancer, chronic lower respiratory disease, Alzheimer�s, and diabetes being some of the leading causes of death. According to the CDC, roughly three out of four adults have at least one chronic disease. I developed and evaluated a multiclass machine learning pipeline to predict U.S. Census regions from state-level chronic disease mortality profiles derived from public datasets. The analytic set included 204 state-year observations (51 jurisdictions, years 2011, 2013, 2015, and 2017) and a four-class target (Northeast, Midwest, South, West). Five classifiers were tuned with Bayesian optimization (XGBoost, support vector machine, random forest, logistic regression, and multilayer perceptron). To reduce leakage risk, we used state-grouped holdout splitting (no train-test state overlap), group-aware cross-validation, and fold-safe preprocessing with scaling and SMOTE applied within each training fold. Model performance was moderate on unseen states: accuracy ranged from 0.575 to 0.700, macro F1 from 0.436 to 0.589, and macro one-vs-rest AUC from 0.828 to 0.916. These findings indicate that a correlative signal between statewide chronic disease mortality profiles and geographic region is present but variable. Future work should expand feature coverage, test external generalization, and utilize explainers before drawing strong policy or clinical inferences.
Description
Date
1/1/2026
Student Status
Junior (Graduating in 2027)
Journal Title
Journal ISSN
Volume Title
Type of presentation
Poster
Collections
Research Projects
Organizational Units
Journal Issue
Citation
DOI
Advisor(s)
Department
Program/Major
Neurobotics and Artificial Intelligence
College/School
College of Arts and Sciences
Organization
Research Category
Mathematical Science