A hybrid approach to segment-type geocoding of New York City traffic data

Abstract

Many types of traffic data are often recorded as (Main street, From street, To street) triples. All the segments between the intersection pairs of (Main street, From street) and (Main street, To street) need to be geocoded, with additional directional constraints. We term this new type of geocoding task as segment-type gecoding in contrast with classic geocoding that takes a street address or intersection and converts the address to a pair of coordinates. Most of the existing geocoding software does not have the capability to handle such segment-type geocoding. This motivates us to develop algorithms and programs for the new type of gecoding in the context of the Effectiveness of Traffic Calming study sponsored by the New York City Department of Transportation (NYCDOT). Due to the level of complexity of the New York City street network, we have adopted a hybrid approach. The hybrid approach includes several algorithms to automatically geocode well-formed traffic records and a software tool built on top of ESRI ArcMap to facilitate manual geocoding of ill-formed traffic records. The hybrid approach has achieved desired accuracies with reasonable manual involvements. We believe the approach is applicable to similar projects that involve segment-type geocoding tasks.