NCJ Number
251177
Date Published
October 2017
Length
9 pages
Annotation
This report presents the features of one of the winning entries in the National Institute of Justice's (NIJ's) "Real-Time Crime Forecasting Challenge," which aimed "to harness the advances in data science to address the challenges of crime and justice."
Abstract
Participants in the challenge were required to identify a forecast area within Portland, Oregon, where certain types of crime were most likely to occur. NIJ provided historical crime data for the Portland police district for the period January 1, 2013 through December 31, 2016. Participants could use this data and any other available data to make their predictions. Entries were judged based on measurements of actual crime data for a 3-month period beginning in March 2017. The co-winner who presented the plan described in this report entered under the category of a "small team/business" (himself alone). The primary data provided by NIJ were records of each call for service (CFS) received by the Portland police district during the period January 1, 2013 to December 31, 2016. Each CFS record included a location (x,y coordinates), crime category and subcategory, a date, and some other information. This paper describes how the co-winner in his category, Warren L.G. Koontz, used tools provided by MATTLAB to analyze the historical data and determine an area within the Portland police district that seemed most likely to have the highest rate of a particular crime (Koontz chose burglary). After describing the data provided by NIJ, this paper discusses the evaluation criterion, the analysis approach, the implementation and test, and the entry submission and results. 3 figures, 1 table, and 3 references