Module 1 began the first focus in GIS Applications with the various ways to conduct crime analysis. I learned many methods including aggregated/grid-based thematic mapping, kernel density mapping, and Moran's I analysis. These three methods were the three I was challenged to conduct in this module's lab assignment.
I was given a few features to manipulate for assignment purposes. These features included a point features of total homicides in Chicago in 2017, a census tract map of Chicago, a grid mask spanning over Chicago, and a boundary shapefile of the greater Chicago area. This boundary was used to set my environment in ArcGIS Pro for all three types of analysis. Using these features, I was to conduct a crime hotspot analysis for all three aforementioned methods for 2017. I was then to compare the three methods and how effective they would be to a police department in predicting future crime using a feature containing points for all homicides in 2018.
The first hotspot analysis I was tasked with was grid-based thematic mapping. I began by employing a spatial join between the grid cells provided and the total homicides in 2017. This gave me a new field in the grid cell providing a count of total homicides in each cell. I then used Select by Attributes to select all cells with a value greater than 0 and exported the selection as a new feature class. I manually selected the top 20% of these cells (quintile classification) and again made a new feature class from the selection. Then, by adding a new field to this feature and using the field calculator to fill it with a constant variable, I was able to use the Dissolve tool to turn my quintile classification into one singular polygon shapefile. The final result is a grid-based thematic crime hotspot map that I will provide a screenshot of here:
The next analysis method I completed was a kernel-density map. Using the Kernel Density tool I was able to create a raster using the total homicide point feature class. In Symbology, I changed the data classification to only have two break values: Those at or above three times the mean. I used to the statistics feature within Symbology to determine the mean. I then used the Reclassify tool to reclassify the raster based on the new break values. This allowed me to use the Raster to Polygon tool to revert the raster back into a polygon. Finally, I used Select by Attributes to select features from the new polygon with a value of "2" (the value I reclassified earlier as areas with a kernel density at or above triple the mean) and exported the selection to a new feature class. The result was a kernel-density map which I will provide a screenshot of here:
For my local Moran’s I analysis, I began with a spatial join between the provided census tract feature and the 2017 homicide feature. This gave me a new field called Join_Count within the census tract feature that provides a count of homicides per tract. I calculated crime-rate by using Calculate Field within the census tract attribute table and used the following equation: [Join_Count / total_households]*1000. This generic crime rate equation presents the number of homicides per 1000 housing units. Next, I altered the Symbology for the census tract feature to show graduated colors, this way I could use Advanced Symbology to employ data exclusion for crime rates of 0. I then used the Cluster and Outlier Analysis tool. I used the updated census tract feature class as the input feature class and my calculated crime rate field as the input field, with all other parameters default. Next, to isolate High-High clusters, I used Select by Attributes. With the HH clusters selected, I exported the selections as a new feature class. Finally, I made the HH clusters into one polygon using the Dissolve tool. I will provide a screenshot of the result here:
To conclude the assignment I evaluated each method based on the total area each hotspot covered, the number of total 2018 homicides that fell in each hotspot, and the crime density of 2018 homicides per square mile of the 2017 hotspots. I determined that the Kernel Density hotspot would be the most useful map in this scenario for a few reasons. First, about 50% of the homicides in 2018 occurred within the 2017 hotspot. This is also true for the Moran’s I hotspot, however the Kernel Density map is superior in that it covers less area. This is important as far as resource allocation and the fact that the Kernel Density hotspot resulted in a higher crime density analysis.
Comments
Post a Comment