Creating open source composite geocoders: Pitfalls and opportunities
Christopher Prener, Branson Fox
Abstract
Abstract Geocoding, the process of converting a street address into a pair of coordinates that facilitate mapping, is a perennial challenge for researchers. Both technical and practical barriers exist to geocoding itself and creating reproducible workflows that string multiple geocoders together (so‐called “composite” geocoders). This article presents a suite of open‐source tools designed to standardize American addresses and geocode them. These tools culminate in a composite geocoder for the city of St. Louis, Missouri, which serves as a proof‐of‐concept for developing sophisticated geocoding processes. We test this geocoder on a data set of 1,822 homicides for St. Louis between 2008 and 2018, comparing the results of our tools to a range of commercial providers. We find varying levels of efficiency and accuracy across the available commercial platforms and demonstrate that our in‐house tools provide high positional accuracy with significant efficiency increases.