Document Type
Thesis
Date of Award
12-31-2024
Degree Name
Master of Science in Data Science - (M.S.)
Department
Data Science
First Advisor
Mengjia Xu
Second Advisor
Hai Nhat Phan
Third Advisor
Mengnan Du
Fourth Advisor
Lijing Wang
Abstract
The misuse of stimulant prescription medications poses a significant and escalating public health concern in the United States, particularly among young adults. Addressing this issue requires sophisticated methodologies capable of uncovering complex patterns and relationships in data. Geometric Deep Learning, a paradigm designed to analyze data with non-Euclidean structures, has achieved remarkable success across various domains, offering a powerful framework for tackling complex graph structure data challenges.
This study leverages Graph Convolutional Networks (GCNs) to predict the likelihood of stimulant medication misuse using data from the National Survey on Drug Use and Health (NSDUH). Individuals are represented as nodes in a graph, while their relationships form edges, capturing social, behavioral, and contextual factors. Experimental results demonstrate the superior performance of the GCN model, achieving 96.1% accuracy, 92.93% precision, and 96.40% recall. can you add the precision and recall too? when compared to traditional machine learning approaches -- Support Vector Machines, k-Nearest Neighbors, Multi-Layer Perceptron, Random Forests, Gaussian Naïve Bayes, and Logistic Regression. The findings underscore the transformative potential of GNNs in predicting stimulant misuse and guiding the design of data-driven prevention strategies. To enhance interpretability, the study integrates GNN Explainer, a tool designed to identify feature importance and explain the model's predictions. By analyzing the critical features influencing misuse predictions, this approach provides great insights into the 20 key factors driving stimulant misuse.
Recommended Citation
Razavi, Hamid, "Advancing prediction of stimulant medication misuse through graph representation learning" (2024). Theses. 2949.
https://digitalcommons.njit.edu/theses/2949
