Flag and boost theories for hot spot forecasting: An application of NIJ’s Real-Time Crime forecasting algorithm using Colorado Springs crime data

AuthorYongJei Lee,SooHyun O
Published date01 March 2020
DOI10.1177/1461355719864367
Date01 March 2020
Subject MatterArticles
Article
Flag and boost theories for hot spot
forecasting: An application of NIJ’s
Real-Time Crime forecasting algorithm
using Colorado Springs crime data
YongJei Lee
School of Public Affairs, University of Colorado Colorado Springs, USA
SooHyun O
School of Criminology, Criminal Justice and Strategic Studies, Tarleton State University, USA
Abstract
By operationalizing two theore tical frameworks, we forecast cr ime hot spots in Colorado Springs. Fi rst, we use a
population heterogeneity (flag) framework to find places where the hot spot forecasting is consistently successful over
months. Second, we use a state dependence (boost) framework of the number of crimes in the periods prior to the
forecasted month. This algorithm is implemented in Microsoft Excel
, making it simple to apply and completely
transparent. Results shows high accuracy and high efficiency in hot spot forecasting, even if the data set and the type
of crime we used in this study were different from what the original algorithm was based on. Results imply that the
underlying mechanisms of serious and non-serious crime for forecasting are different from each other. We also find that
the spatial patterns of forecasted hot spots are different between calls for service and crime event. Future research should
consider both flag and boost theories in hot spot forecasting.
Keywords
Crime hot spot, forecasting, population heterogeneity, state dependence, Excel
Submitted 03 Nov 2018, Revise received 03 May 2019, accepted 10 Jun 2019
Introduction
Over the past two decades, hot spot policing has emerged
as one of the most effective strategies in reducing crime
at place. A large number of experimental and review
studies have found that allocation of police resources to
hot spots significantly reduces crime (Braga, 2005). Typi-
cally, hot spot detection has been widely used in police
practice as an alternative way to predict future crime
places as it assumes that previous crime places are more
likely to experience future crime. These hot spot detec-
tion methods use cross-sectional data from the most
recent period in two-dimensional (in both time and space)
clustering methods (Gorr and Harries, 2003; Gorr and
Lee, 2012, 2015) or using kernel density smoothing
(KDS) (see Eck et al., 2005).
However, missing from this hot spot detection practice
are approaches and methods that deal with hot spot fore-
casting. Hot spot forecasting requires more sophisticated
methods and computational power relative to hot spot
detection because it deals with places where crime has not
occurred before. However, with the advent of computer-
aided mapping and geographic information systems,
employing hot spot forecasting methods became more pos-
sible (Bowers et al., 2004; Cohen et al., 2007; Gorr and
Olligschlaeger, 2001, 2003; Groff an d La Vigne, 2002).
Corresponding author:
YongJei Lee, School of Public Affairs, University of Colorado, 1420 Austin
Bluffs Parkway, Colorado Springs, CO 80918, USA.
Email: ylee@uccs.edu
International Journalof
Police Science & Management
2020, Vol. 22(1) 4–15
ªThe Author(s) 2019
Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/1461355719864367
journals.sagepub.com/home/psm

To continue reading

Request your trial

VLEX uses login cookies to provide you with a better browsing experience. If you click on 'Accept' or continue browsing this site we consider that you accept our cookie policy. ACCEPT