Data analytics of fatal police shootings involving individuals in mental health crisis

Abstract

In this study, a Monte Carlo simulation model was implemented to determine if having a licensed clinical social worker who responds to mental health emergencies would reduce the number of individuals with severe mental illness who are fatally shot by on-duty police officers. Factor analysis was performed to determine correlation and remove highly correlated features. This was followed by a logistic regression model that explored relationships between the remaining features and the target variable: whether the individual killed showed signs of mental illness prior to being shot. A self-organizing map was used for clustering as well as for feature selection. This method of unsupervised feature selection provided four features that were used to run the optimization and simulation models: the use of a body camera, if an individual was armed, the violent crime statistics for the state, and the reported threat level of the individual towards police officers. First, the dual annealing optimization was performed with an original objective function, then a utility function of a licensed clinical social worker was added. The results showed that increased body cameras and the added utility function helped reduce the number of fatal shootings involving individuals with severe mental illness. The Monte Carlo simulation was performed first, with the original objective function, and second, with the added utility function. This confirmed the results of the optimization algorithm that increased body cameras and a licensed clinical social worker could reduce fatal police shootings of individuals with severe mental illness.