Dana Steil Harding University Branch, Center for Advanced Public Safety

Transcription

1 Multi-Factor Optimization of Police Presence Using An Automated GIS Tool Dana Steil Harding University Branch, Center for Advanced Public Safety

2 Outline User s Perspective Inspiration from DDACTS Delivery Channel Target Selection Target Presentation Demonstration Technical Perspective Data Sources/Storage Application Architecture Target Criteria Target Selection Algorithm Target Compression for Transfer to client Target Rendering

3 Inspiration from DDACTS DDACTS (Data Driven Approach to Crime and Traffic Safety) ideals from NHTSA What data do we have? What would be informative/useful? How do we make it accessible (technically and practically)?

4 Delivery Channel ADVANCE - Alabama Dashboard for Visualization Analysis and Coordinated Enforcement

5 Target Selection

6 Target Presentation

7 Demo

8 Data Sources

9 Application Architecture Microsoft MVC Web Application Web services SQL Server SDE Databases

10 Target Criteria Model Current User List of reports The user can select from an array of reports and can combine them to overlay various targets. Report Label Color (for target cells) List of criteria Reports combine criteria so that cells are only targets if they satisfy each criterion in the list. Criterion Threshold value Threshold operator Datasource type Filter Each criterion filters cells that are not targets.

11 Target Selection Algorithm Problem: Given a set of criteria, how do we decide which cells are targets? 1 st Score cells by counting the number of events associated with each criterion that occurred within the cell. 2 nd Compare each cell s score to each criterion s threshold value. The threshold value is the average score for each cell multiplied by a scalar provided by the user. Examples: 2 times average, 1/2 average

12 Target Selection Algorithm All cells are assumed to be targets until eliminated by a criterion. If the cell satisfies the criterion s threshold value, it is a target for that criterion. A high criterion is satisfied by cells that exceed its threshold value and a low criterion is satisfied by values less than its threshold. A cell is target in a report if it is a target for each criterion that the report contains. Example: High Crash and Low Citation cell is a target iff it satisfies the High Crash criterion and the Low Citation criterion.

13 Need for Target Compression Each cell is.25 sq mi Bounding box for AL is 680 mi by 420 mi 285,600 cells

14 Target Compression Suppose the black cells in this table are hot. This could be represented in a Boolean fashion as However, on maps with a large number of cells this results in a huge amount of data if stored as True/False values. Instead, we can represent this table as the integer number equivalent to the above binary representation: We can divide the map into blocks of 32 cells and represent them with these numbers. We then return these numbers to the client in the form of a commaseparated string, leaving blank blocks with no hot cells.

15 Target Rendering Client receives this compressed data as string of comma separated numbers Comma-separated string is converted back to its binary form, (32 cells each). Consecutive target cells at the are rendered as a single polygon ring.

16 Thank You! Dana Steil Phd Assistant Professor of Computer Science Harding University Center for Advanced Public Safety

17 Abstract As resources continue to shrink for public safety, data-driven policing strategies are dramatically increasing in popularity. DDACTS (Data-Driven Approaches to Crime and Traffic Safety) is a discipline designed to determine the optimal locations for police presence based on the multiple factors of previous criminal activity and vehicle crashes. The idea in DDACTS is to determine map sectors (e.g., square mile blocks) where crashes and crime are prevalent, and provide increased police presence in these sectors. We have developed a GIS-based tool to automate this strategy by overlaying high crime with high crash locations. In this presentation, we describe the design and implementation of this tool, and illustrate its use with actual police historical data in an Alabama city.