The Florida Department of Transportation (FDOT) has implemented freeway service patrol programs at major urban areas to ease traffic congestion due to non-recurring incidents. This paper describes an effort to use data mining techniques to extract information from the service patrol databases for Broward and Palm Beach counties, Florida. A pattern-based geocoding process was developed to geocode the databases. Geocoding was needed because the incident records do not contain specific geographic coordinates, but only a general description of the incident locations. An intelligent parser program was developed to scan and parse location information patterns to identify specific geographic locations. The program was able to achieve a high geocoding rate of over 95% for the service databases. A spatial analysis example making use of the geocoded data was demonstrated. Copyright ASCE 2006.