Forecasting Crime Using Exploratory, Retrospective, and Prospective Methods
ISBN 978-85-88783-11-9
Authors
1Leitner, M.
1LOUISIANA STATE UNIVERSITY Email: mleitne@lsu.edu
Abstract
This presentation reports on an ongoing research project about the prediction of criminal events in major Austrian cities using exploratory, retrospective and prospective analysis methods. This research project is a collaboration of a multi-disciplinary consortium of several law enforcement agencies (federal, state, and local), a research and technology institute (Joanneum Research, http://www.joanneum.at/), the private sector (SynerGIS, http://www.en.mysynergis.com/), the Department of Geoinformatics – Z_GIS at the University of Salzburg (http://www.zgis.at/), and the Institute for the Sociology of Law and Criminology (http://www.irks.at/). It is funded by the Austrian Research Promotion Agency (https://www.ffg.at/en) as part of its Austrian Security Research Program KIRAS. This research primarily applies geospatial methods and technologies to analyze the probability of future crime occurrences. This study is also the first to investigate whether predictive concepts and methods developed in the US and the UK over the past decade can be successfully applied to Austria and, by extension, to other countries of mainland Europe that share a similar history, culture, economy, etc. This may also be the most comprehensive research to date on this topic. The results of this research are both immediately shared with the police, where they are implemented when deemed relevant and practicable, and also serve as the main input into the building of a prototype software tool for law enforcement agencies to predicting crime. The study areas include the three largest cities in Austria and two medium-sized cities. The crime data are collected from the Security Monitor (SIMO) administered by the Austrian Federal Police that stores all reported crimes in Austria since 2004. The main attributes for each reported crime include the exact location of the crime occurrence (the x- and y-coordinates of the address, if known by the victim), the time of the occurrence (to the minute, if known by the victim), and the crime type. Non-crime data sources are numerous and constitute, for example, data from the federal government, such as Statistics Austria and Geographic Information System (GIS) databases from individual states that include the selected cities. In addition, the results from a recent and already completed KIRAS project led by Joanneum Research that identified the most significant criminogenic factors of crime in Austria have also been incorporated. The analysis thus far has been both exploratory and confirmatory applied to both retrospective and prospective analysis. The exploratory analysis has, for instance, found no statistically significant relationship between apartment burglaries and full moon days and a significant peak in apartment and home burglaries in specific neighborhoods in Graz during the fourth week of January from 2009-2013. This information led the police to concentrate resources in these neighborhoods during the fourth week in January 2014, resulting in an above average clearance rate of these crimes. The retrospective analysis has focused on the application and evaluation of spatial (Getis-Ord Gi*, local Moran’s I, kernel density estimation, spatial and temporal analysis of crime, and nearest neighbor hierarchical clustering-NNHC statistics) and spatial-temporal hot spot methods (near repeat calculator-NRC, space-time Gi*, and SatScan statistics). Of the purely spatial hot spot methods, the NNHC statistic has shown the most promising results thus far. However, no conclusive evidence has been found in the evaluation of the three spatial-temporal statistics selected. Already completed research has confirmed both the near-repeat phenomenon for robberies and burglaries in Vienna and the correct spatial-temporal prediction of between 25-50% of future robberies, auto-thefts, and burglaries with the risk terrain modeling approach (a prospective analysis method) for the city of Salzburg, Austria.
Keywords
Crime forecasting; Crime hot spots; Austria