Paper Type
Complete
Abstract
This research project presents a novel application of a Deep Neural Network (DNN) to predict drug overdose risk using social media data, addressing traditional public health surveillance limitations. A Design Science Research (DSR) framework and randomized experimental design underpin the development of an AI artifact that processes georeferenced Twitter (X) data for features such as drug references, sentiment, cognition, and polarity. As an independent variable, quantitative evaluation shows that location significantly improves predictive performance. The DNN model achieves strong results for drug presence prediction (MAE = 0.0404 and RMSE = 0.1411). Geospatial visualizations reveal close alignment between predicted hotspots and actual tweet distributions. Findings confirm that real-time, community-level overdose risk prediction using social media language and deep learning is feasible. This research advances AI-enabled Spatial Decision Support Systems (SDSS) for public health, offering a scalable, rapid-response framework to enhance overdose surveillance and inform targeted interventions.
Paper Number
2217
Recommended Citation
Corso, Anthony; Asariah, Jason S.; Park, Ki-Pung; Corso, Nathan A.; and Lee, Esther Lee, "A DSR Approach for Geospatial AI: Prediction of Drug Overdoses" (2025). AMCIS 2025 Proceedings. 27.
https://aisel.aisnet.org/amcis2025/intelfuture/intelfuture/27
A DSR Approach for Geospatial AI: Prediction of Drug Overdoses
This research project presents a novel application of a Deep Neural Network (DNN) to predict drug overdose risk using social media data, addressing traditional public health surveillance limitations. A Design Science Research (DSR) framework and randomized experimental design underpin the development of an AI artifact that processes georeferenced Twitter (X) data for features such as drug references, sentiment, cognition, and polarity. As an independent variable, quantitative evaluation shows that location significantly improves predictive performance. The DNN model achieves strong results for drug presence prediction (MAE = 0.0404 and RMSE = 0.1411). Geospatial visualizations reveal close alignment between predicted hotspots and actual tweet distributions. Findings confirm that real-time, community-level overdose risk prediction using social media language and deep learning is feasible. This research advances AI-enabled Spatial Decision Support Systems (SDSS) for public health, offering a scalable, rapid-response framework to enhance overdose surveillance and inform targeted interventions.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.
Comments
IntelFuture