Module 4: Crime Analysis

This module introduced the application of GIS to crime analysis. I explored different methods of crime hotspot mapping. We covered grid-based mapping, local Moran's I, and kernel density in depth. In the lab, I used those techniques to make three maps of 2017 Chicago crime.

To make the grid-based thematic map, I did a spatial join of the ½ mile gird and the 2017 homicides saving it as homicide_grid. I then selected by attribute to select all the grids with a homicide count and made those a new feature class. Next, I manually selected the top 20% of homicide grid cells in the attribute table. I saved this as a new file. After creating a new dissolve field, I converted the feature class into a single polygon. I did this by using the dissolve tool with the inputs provided in the instructions.

To make the kernel density map, I used the provided inputs in ArcPro’s kernel density tool. I then edited the symbology. I changed the symbology to only two break values, 8.64 and 38.92. I just these same values to reclassify the raster. Next, I converted the kernel density to a polygon and selected only the “2” values (representing those above 3x the mean). I exported the selection and made a new features class in the same method as the previous map.

I made the local Moran’s I map by first doing a spatial join of the census tracts and 2017 homicides. I added a crime rate field to the newly created feature class. I used the field calculator to determine how many homicides were in each tract per 1,000 houses. I used the cluster and outlier analysis tool with the joined feature class and crime rate field as inputs. This made the local Moran’s I map. For this map I selected the HH clusters, saved those a new feature class, and dissolved it into one polygon.

I believe the most useful map to a police chief would be the kernel density map. This is because this hotspot mapping technique is based on the point data directly. The other maps used a 1/2 mile grid and census tracts as inputs. Using the kernel density map, the chief would be able to target hotspots within the other maps’ grid cells. The kernel density has a similar percentage of 2018 homicides in 2017 hotspots compared to local Moran’s I. It also has a higher crime density than local Moran’s I. I think the kernel density provides the best balance of those two measures of all three hotspot mapping techniques.