Effects of Data Quality on Predictive Hotspot Mapping

• NCJ Number: 239861
• Author(s): Timothy C. Hart Ph.D.; Paul A. Zandbergen Ph.D.
• Date Published: September 2012
• Length: 89 pages
• Grant Number(s): 2009-IJ-CX-0022
• Agencies: NIJ-Sponsored
• Publication Type: Research (Applied/Empirical)
• Award: Effects of Data Quality on Predictive Hotspot Mapping

Annotation

This final NIJ technical report presents the results of a study examining the effects of data quality on predictive hotspot mapping.

Abstract

This study had three primary objectives: 1) determine the empirical descriptions of the quality of a range of typical geocoding techniques employed in crime mapping; 2) characterize the effects of data quality on the robustness of selected predictive crime hotspot mapping techniques; and 3) determine the effects of analysis method, crime type, urban morphology and parameter settings for predictive crime hotspot mapping techniques given a range of typical data quality parameters within the context of the accuracy and precision of hotspot prediction. The major finding from this study on the effects of data quality on predictive hotspot mapping was that geocoding quality was affected by variations in crime type as well as reference data used during the geocoding process. In addition, the study found that no single technique was more accurate than any other, and that the procedures used by analysts were highly influenced by many factors. Based on this finding, the authors of this study developed the following set of recommendations for use by researchers and analysts engaging in hotspot analysis: 1) assess the overall quality of input address information prior to geocoding; 2) disaggregate crime incidents and geocode like crime events separately; 3) tailor geocoding procedures to fit specific needs; 4) geocode to local street centerline reference data, if it is available; and 5) characterize positional accuracy prior to additional analysis. Data for this were obtained from the analysis of over 400,000 crime incident records from 6 large law enforcement jurisdictions in the United States. Study limitations are discussed. Figures, tables, references, and appendixes