Document Type
Report
Publication Date
5-2022
Abstract
Given recent events in Buffalo and Uvalde, can we predict where targeted violence, specifically active shooter attacks, is most likely to occur? Building on our 2021 report, we re-examine the risk of targeted violence – as measured by mass shooting incidents -- across the United States. In our original report, we examined whether Machine Learning (ML) techniques could help forecast the location of domestic extremist hate groups and targeted violence. We found that basic ML algorithms could accurately forecast hate group operations, but the location of mass shootings was relatively stochastic. In other words, communitylevel risk indicators poorly predicted where a mass shooting was likely to occur. At the time, we noted the model’s poor forecast accuracy could potentially be due to the 2020 COVID pandemic, which generated a sharp reduction in mass shooting incidents due, in part, to stay at home orders. Since a ML model learns from historical reporting to forecast future trends, an outlier year could erroneously cause the model to look worse than it is. Since 2020, the FBI reported active shooter incidents increased 52% in 52%, suggesting a return to prior levels. In practice, this means we have better data for re-evaluation.
Recommended Citation
National Counterterrorism Innovation, Technology, and Education Center and Malone, Iris, "Predicting Targeted Violence: An Update" (2022). Reports, Projects, and Research. 29.
https://digitalcommons.unomaha.edu/ncitereportsresearch/29