An abundance of data confirms a national opioid epidemic, but much of this data is analyzed and reported in retrospect. This project addresses that limitation for one aspect of the problem, specifically where are opioid hotspots currently emerging, and where are they likely to emerge in the future? The project developed an extensible Criminal Predictive Analytic Platform (CPAP) to gather a large number of publicly available data sets in order to predict geographic areas of future opioid abuse. The team developed a data processing pipeline using open source tools to correlate known data sets (trended and geographically-attributed historical opioid fatalities) with a multitude of historical open source intelligence (OSINT) data sets. The team then trained a machine learning classification model to predict future opioid fatality rates using near real-time current OSINT sources.
The researchers developed an interactive graphical map application to present the resulting information to users. These users, who may include but are not limited to investigators and law enforcement, can view the data at a national level, but also have the ability drill down and view details for specific geographic areas and time windows. The application and results have been demonstrated to DHS components working different aspects of the opioid problem. The researchers designed the data processing pipeline, algorithms, and application to support the addition of new data sources so that the system can be used and enhanced by law enforcement and others working to fight the epidemic.