Geocoding has been a widely used technology in daily life and scientific research for at least four decades. Especially in scientific research, geocoding has been used as a generator of spatial data for further analysis. These uses have made it extremely important that geocoding results be as accurate as possible. Existing global-weighting approaches to geocoding assume spatial stationarity of addressing systems and address data characteristic distributions across space, resulting in heuristics and approaches that apply global parameters to produce geocodes for addresses in all regions. However, different regions in the United States (US) have different values and densities of address attributes, which increases the error of standard algorithms that assume global parameters and calculation weights. Region-based dynamic weighting can be used in probabilistic geocoding approaches to stabilize and reduce incorrect match probability assignments that are due to place-specific naming conventions which vary region-to-region across the US. This study tested the spatial accuracy and time efficiency of a region-based dynamic weighting probabilistic geocoding system, as compared to a set of manually corrected geocoding results within Los Angeles City. The results of this study show that the region-based dynamic weighting probabilistic method improves the spatial accuracy of geocoding results and has a moderate influence on the time efficiency of the geocoding system.
