Applications in GIS Module 4

This week we learned how to produce hotspot maps using a three different methods. We used crime data from different cities to come up with different hotspot maps that either showed what areas were experiencing certain types of crime or where a certain type of crime was predicted to occur. In order to carry out the analysis we used grid based hotspot mapping, kernel density hotspot mapping and finally Local Moran's I hotspot mapping. Below are the hotspots maps that I produced along with a brief description of the processes involved in carrying out the analysis. 

This is the data that was produced from the grid based hotspot mapping process. This map represents partial homicide data within the Chicago area in 2017. Because this analysis is based on initial grids the results came out in nicely shaped polygons. This analysis required spatially joining data to attach homicide number to the grids. Once the data was joined, certain attributes were selected and the data was filtered to show only the areas with the highest homicide rates. The dissolve tool was used to clean up the data by eliminating the resulting multipart polygons. This data produced in this analysis was very close to the method which was determined to the best at predicting future homicides. This method was ranked second in effectiveness because it was slightly less precise in identifying problem areas. 

The second method we used was the kernel density hotspot mapping method. The kernel method required more steps than the grid method. The kernel analysis produces a raster with associated numerical values. After adjusting the data to better represent the data the raster was reclassified and converted to a shapfile. The resulting shapefile resulted in the largest geographic area of the three methods we used. Similarly the kernel method also resulted in the most homicides predicted. This method is likely not the preferred method as it spreads the data over a large geographic area. 

The final method we used to create hotspot maps was the Local Moran's I method. This method considers areas and occurrences to derive the final product. The methods used are similar to the kernel density method with the exception of using different more complex analysis for the Moran's method and it also required a bit more data manipulation. I believe this method to be the most precise of the methods we tried. This method and the grid method had an equal amount of homicides projected, however the Moran's method had the smallest geographic area. A smaller geographic area means the numerical data produced should be more precise or at least more relevant to a given area. The Moran's method is likely more precise because it considers more factors and takes into account surrounding areas when assembling a shape.