Authors: Zhiyue Xia*, Center for Geospatial Information Science, Department of Geographical Sciences, University of Maryland, Kathleen Stewart, Center for Geospatial Information Science, Department of Geographical Sciences, University of Maryland
Topics: Geography and Urban Health, Spatial Analysis & Modeling
Keywords: drug-related crimes, heroin, synthetic drugs, fentanyl, random forest, machine learning
Session Type: Virtual Paper
Start / End Time: 3:05 PM / 4:20 PM
Room: Virtual 8
Presentation File: No File Uploaded
Substance use disorder affects millions of adults in the United States and is a public health concern that impacts every state in the nation. For this research, we examined spatial and temporal patterns of drug-related crimes and detected clusters of crime incidents involving particular types of drugs to distinguish hot zones where drugs are likely to present that in turn can provide a basis for assessing and providing drug-related services. We investigated spatial and temporal patterns of more than 52,000 drug-related crime incidents in Chicago, IL between 2016 and 2019, focusing on two different drug types, heroin and synthetic narcotics, that have been key drugs during the opioid crisis in the U.S. Our analyses showed that crimes involving heroin were clustered in two hotspots in 2016 that diminished in size to a single hotspot in 2019, while the pattern of synthetic narcotics evolved from a scattered pattern in 2016 to two distinct and larger hotspot areas by 2019. We designed a random forest machine learning model to identify what key locations, built environment and sociodemographic factors were correlated with these patterns, and establish the top contributing factors that were estimated to be important for identifying drug hot zones over time. Spatial and temporal autocorrelation was accommodated in the machine learning model that ingested data from multiple geospatial data sources to investigate patterns of drugs and their associated spatiotemporal characteristics in an urban setting.