U.S. flag

An official website of the United States government, Department of Justice.

NCJRS Virtual Library

The Virtual Library houses over 235,000 criminal justice resources, including all known OJP works.
Click here to search the NCJRS Virtual Library

Predicting Crime Scene Attendance

NCJ Number
221862
Journal
International Journal of Police Science & Management Volume: 9 Issue: 4 Dated: Winter 2007 Pages: 312-323
Author(s)
Richard Adderley; John W. Bond; Michael Townsley
Date Published
2007
Length
12 pages
Annotation
This British study examined the feasibility of using a data-mining computer model to predict which crime scenes were more likely to offer the best opportunity for recovering forensic evidence samples, such as fingerprints or DNA, thus facilitating prioritization for the deployment of crime-scene investigators (CSIs).
Abstract
The computer model had an accuracy of 68 percent in predicting which crime scenes would yield fingerprints or DNA samples for analysis. This was significantly better than the accuracy of the judgments of a sample of CSIs regarding which crime scenes were most likely to yield evidence for forensic analysis. The authors caution, however, that the results of the computer model should not be used in isolation. Other factors that should be considered in deploying CSIs to crime scenes include the vulnerability of victims, repeat victimization, the physical location of CSIs, and police agency policy. Suggestions are offered for further research. The study used data from the Northamptonshire Police Force (England), which records all cases into an ORACLE-based relational database designed and implemented in-house. Crime and forensic data obtained between January 1, 2000, and July 19, 2005, were used for the study. The datasets were merged to produce 28,490 individual records related to volume crime scenes: burglary in dwellings, burglary in commercial buildings, and theft of and from motor vehicles. The dependent variable was the forensic evidence collected. The independent variables were subdivision, beat, sub beat, 500-meter Ordinance Survey grid reference block, offense type, and month of the crime. Two supervised learning algorithms were used to model the data: classification neural network and Naive Bayes. 3 tables and 28 references