期刊名称：GI_FORUM - Journal for Geographic Information Science
电子版ISSN：2308-1708
出版年度：2015
页码：30-40
DOI：10.1553/giscience2015s30
出版社：ÖAW Verlag, Wien
摘要：This research project deals with the implementation and evaluation of the Risk TerrainModeling (RTM) technique, which allows localizing places, where the probability is highthat a crime event will take place. RTM does not focus on previous events that happened,but on risk factors, which have an influence on the environment and can increase the probabilityof the risk that a crime will be committed. RTM is a recently developed approach,that has not yet been tested in Austria. Using the example of the city of Salzburg, predictionsare made for the crime events assault, auto theft, burglary, and robbery for 2013 and2014. In addition, the results of 2013 are evaluated and compared. Using the RTMDx Utilitysoftware, risk factors that correlate with the crime event as well as their influence can beidentified. Based on these results, the risk factors can be operationalized to risk map layers,which is done using two models developed within ArcGIS. After that, the risk map layersare combined to a final risk terrain map, which is classified and finalized according to cartographicaspects. The evaluation for the predictions of 2013 is done using the PredictiveAccuracy Index (PAI), based on a model developed in ArcGIS. The results of the evaluationand the percentage of correctly predicted crime events in respect to the size of the predictedareas are shown. In sum, 27 models were calculated and predictions made, becausecrime events have been partly separated into seasons or sub-types. The best predictive accuracyis reached for assaults, which includes results for the seasons spring and summer andfor robberies, with PAI values of 31, 23, and 18, respectively. In contrast, the predictionsfor burglaries and auto thefts performed rather poorly, with PAI values between two andfour. Overall, the RTM technique can be applied to Austrian cities, although the accuracyof the predictions varies. Additionally, the availability and quality of the risk factor data arecrucial for the accuracy of the predictions.
