Crime Analysis: Heat Maps In Action

 Crime Analysis in GIS

Specifically using Esri software suite such as Arc GIS Pro to build crime heat maps to display significantly high areas of crime. This can be based on crime rates using population data or area data, or even using kernel density calculations to define local clusters of crime.  

Different algorithms such as Moran's I can define grid like hot spot areas of significant crime at a 95% confidence interval, allowing you as a GIS analyst to inform your command staff at your local Police Department  the areas in which Patrol should focus their efforts. 

Defining a Crime Rate

This map below is an example of calculating crime rate of a specific type of crime (Burglaries) in a specific area (Washington D.C.) using a standardization of population. We are calculating the number of Burglaries per 1,000 Housing Units. We started out using SQL Expressions and Table joins to select Burglaries from our crime point data, then joining the census data of population and housing units to our crime table. Then we calculated the crime rate of join counts / housing units, times 1,000. Then we excluded 0 data from our symbology and edited the advanced settings to define our classification intervals (burglaries_tract.crimerate = 125 OR burglaries_tract.crimerate = 1500). This excludes the outliers from our data. 

Burglaries Per 1,000 Housing Units - Washington D.C.

Things to know...

In Crime Analysis, most GIS professionals track percent change of each major crime type, such as Armed Robbery or Homicide as a YTD calculation, 28 day change, 14 day change, 6 month change, etc...

This is a great way to track changes in criminal activity of a police precinct or jurisdiction. Using an arc GIS Dashboard to map these changes and track the statistics is a must for most crime analysis Units. 

What is a crime rate?

Crime Rate - 

This is important to track the crimes per population in an area. If we just used crime numbers, then we would not have accurate data. We must standardize the crime data with population, or even acreage of areas to normalize it. If you had 10 murders in an are of 100 people, that would be significant. But if you had 10 murders in an area of 100,000 then it would be not as significant. 

Crime Density - Calculating the density of crime. We do this in many ways, such as using  kernel density to map local clusters of significantly high crime areas based on crimes per square mile or kilometers. These areas have a confidence interval of 90% and sometimes higher of violently high crime areas. 

Hot Spots: Confidence Intervals, Z-Scores, Mean Max, Min, etc...

In crime analysis we identify hot spots for the command staff, patrol Lieutenants, and our police officers so that they can identify the growing crime threat and build a tactical plan to deter it. Looking at areas above the average crime, or high Z-score Percent changes are great ways to do this. By mapping these areas, awe can identify them and enforce the law appropriately. 

Kernel density Heat Maps: Local Clustering

At my current place of employment (Athens-Clarke County Police Department) we use kernel density a lot to map crime. By using local clustering, we can identify crime areas in a heat map that is easy to see and read. In arc GIS Pro, we can use the kernel density tool (located in the Crime Analysis Ribbon) to identify local clustering. 

Select your crime data(in this case we used a SQL expression to identify only Aggravated Assaults) and defined an appropriate cell size of 100 ft, and a search radius of 1,320 ft. The area unit we are using is square miles. Then we exclude the 0 data from our symbology, and set the classes in our symbology to the mean distribution of our data. We just multiplied the mean by a +1 to get 5 times the mean as our last classification value in our symbology. Below is a crime Analysis Heat Map or Hot Spot Map that indicates an extremely high amount of Aggravated Assaults clustered in areas of White and light Purple. 

Assaults In Washington D.C. Using Kernel Density

3 Different Types of Hot Spot Analysis for Chicago Crime Data

Grid-based thematic mapping

In this map, we used grids to aggregate point data within each grid. then we calculated data based on a SQL expression of selecting only homicides from our crime data, and selected the top 20 percent of grids with homicides. 

Kernel density

Using the same homicide data, we ran a kernel density tool and built a heat map of high clusters of crime. We then converted our raster to a polygon file using a reclassification tool, and selected the value of 2 which was our hot spot areas. 

Local Moran's I

Using this type of crime analysis mapping, the Local Moran's I required us to use the same homicide crime data from before and run the Cluster and Outlier Analysis (Anselin Local Moran's I) in Arc GIS Pro. This resulted in spatial clusters that had 4 categories, ranging from low to high. We ran a SQL expression to get only the High High (HH) clusters and exported the data out as our final crime hot spot map. 

Conclusion

Lastly, we calculated the area of each hot spot data layer, and then we calculated how many 2018 crimes were in each hot spot layer. 

This is very common in crime analysis, as we use current or historical heat maps to predict future crime counts and areas. This is known as predictive analysis. We want to be proactive rather than reactive to deter crime. This is the best way in practices to do so. then we calculated the total % of data within each hot sport and the crime density per hot spot based on square miles and the number of homicides in 2018. 

In conclusion, the hot spot map I think that was the most efficient and best practice is the Kernel Density. My chief at the ACCPD prefers Kernel Density maps as they are easy to see, understand, and you canoverlay crime point data with crime symbology over them to give you a solid spoatial reference of the crime areas. But don't take my word for it...

The table above indicates that the kernel density method has a crime rate of 11.9 which is higher than the other two methods. However, it does not have the largest area coverage, and it does not have the highest number of crimes falling within its hotspot coverage. But this means that the kernel is extremely efficient and prepares the detectives in CID and the Patrol Unit of the highest crimes with a more specific area. This is crucial when briefing a chief of police or the command staff, as resources and officers are limited across every department. we want to be accurate, specific, and tailor our response to crime to be as efficient as possible. We cannot protect the entire county all the time. We need to know where to schedule patrol units, where to look, and where we should have a strong presence in the community to combat and deter crime. The kernel density is also extremely fast to use and is great for everyday workflows. I highly recommend it as a crime hotspot reporting tool over the Grid Overlay and Local Moran's I statistic method. 

Check Out this video on Crime Analysis using the Arc GIS Pro Crime Ribbon. The most prominent tools used are optimized hotspots, kernel density, emerging hotspots, tracking gang members and gang territories, and lastly cell phone analysis (this topic is the most crucial in crime investigating)

Don't Forget To Subscribe!