Anomaly Hot Spots
Just-in-Time detection of crime sprees.
Leverages Historical Data
The algorithm looks at 3 years of crime data grouped into 18 crime types.
The Opportunity
Traditional hot spot analyses often identify the same high-crime areas known by seasoned officers, offering little new insight. The Lethbridge Police Service (LPS) needed a tool to detect unusual crime spikes—emerging hot spots that fall outside of these predictable patterns—enabling officers to respond proactively to changing crime trends.
The Approach
Leveraging anomaly detection with techniques like kernel density estimation and z-scores, I developed an Anomaly Hot Spots tool focused solely on identifying new crime spikes and sprees. This approach highlights areas experiencing unusual activity, helping officers locate and address emerging issues before they escalate.
The Impact
The Anomaly Hot Spots tool has transformed how LPS identifies and responds to crime patterns. By surfacing unexpected crime spikes in real time, it empowers officers to proactively address emerging threats, enhancing public safety and enabling smarter resource allocation to areas that need it most.
Features
Spatial Hot Spot identified Using KDE
For every 30 day period over 3 years, Spatial Kernel Density Estimates are used to monitor crime spikes.
Densities are Normalized
KDE results are normalized through a z-scores calculations.
Hot Spot Map Updated Daily
Hot Spots are refreshed daily on the LPS Intelligence Portal.
Technology
- Coded in R
- Libraries used: tidyverse, odbc, sf, leaflet, SpatialKDE, raster, terra and others
- Automated through Node-Red
- MS SQL for audit logging
- Active Directory for authentication
Disclaimer
The demo images presented here are for illustrative purposes only and use fictitious incidents. No real incidents or sensitive data are shown.
