Pattern Recognition in Large Police Crime Data Sets

• NCJ Number: 177222
• Author(s): T O'shea; T Muscarello
• Date Published: 1997
• Length: 15 pages
• Grant Number(s): 95-IJ-CX-0082
• Agencies: NIJ-Sponsored
• Publication Type: Report (Study/Research)

Annotation

Since police departments began to keep records of criminal incidents, they have attempted to put this information to strategic and tactical use and pattern discrimination within the crime data set has always been at the heart of crime analysis.

Abstract

A pattern refers to an individual or a group of individuals who are characterized by the fact that they commit a series of criminal offenses over an extended period of time. Additionally, these individuals choose a particular crime category and habitually execute the same method of operation through a series of separate criminal incidents. Pattern recognition methods of experienced robbery detectives in the Chicago Police Department, specifically artificial neural networks and cluster analysis, are examined. Neural networks are systems that seek via hardware or software to simulate the architecture and working of the human brain. These networks are adept at finding patterns in historical data that can be used for predictive or forecasting purposes. Neural nets must be trained to identify patterns before they can be used to recognize or categorize unknown patterns. Two basic modes of learning associated with neural networks are supervised and unsupervised. Supervised learning procedures have achieved a reputation for producing good results in practical applications and represent the most commonly used learning algorithm. Cluster analysis refers to a range of multivariate statistical procedures, and the method seeks to group cases according to similarities in their defining characteristics. Automated techniques to facilitate the use of cluster analysis specifically and pattern recognition generally are briefly described. 1 note and 2 exhibits