Information-theoretic methods applied to dispatch of emergency services data

Abstract

The dispatch of emergency services is a complex cognitive task. Current decision support methods rely heavily on manual analysis of maps and map overlays. This paper aims to use call for service/emergency (CFS) dispatch data from various cities to look for patterns not usually amenable to visual analysis that could be used to create decision support tools or methods for dispatchers who must allocate first responder resources under emergency conditions. The authors have collected from the Police Data Initiative, a publicly available government repository that contains millions of annotated 911 dispatch records. The authors have selected three major American cities (Hartford, CT; Lincoln, NE; and Orlando, FL). Three experiments are performed to assess possible benefits of augmenting conventional manual methods with automated analysis derived using methods of data science. In particular, high-dimensional and non-linearly coded information not amenable to manual analysis are considered.