Scope

The project covers an area limited to New Braunfels' Extra-Territorial Jurisdiction (ETJ)—a 3.5 mile extended area around the city in which NBPD officers have jurisdiction to serve warrants. The tool is also limited to Class C Misdemeanors.

Methodology

Based on the needs assessment, our primary task consists of constructing a geocoding model that will spatially assign each warrant’s attributes to their respective physical addresses in an efficient and effective manner.

Development of the warrant geocoding model consist of four phases and built using ArcGIS version 10.

1. Phase I: Data Preparation/File Conversion - warrant data will be parsed and batched into a Excel based .csv file using a customized Python script. It will be the responsibility of New Braunfels Police Dept to run the script to batch new warrant data before it is ran through the geocoding tool.

2. Phase II: Basic Warrant Tool Development - the structure of the model was built in Modelbuilder. We used standard ESRI geocoding framework. We determine functionality must meet these minimum criterion.
   
   a. The tool must filter warrants outside of the ETJ.
   b. The model must be built in a configurable file geodatabase.
   c. Hit accuracy at 80%. Ideal over 90%.

3. Phase III: Advanced Tool Construction

   We design the model using file geodatabases and folders to replicate the workspace environment as close as we possible to make it configurable to the ArcSDE environment. Modifications to make this tool accessible in the ArcSDE environment will be configured by IT or GIS personnel.

4. Phase IV: Tool Proofing and Refinement

   The goal for the warrant model is for its integration into the NBPD ArcSDE server whereby it will become available for further development and utilization. Once this is accomplished, the model’s regular output (geocoded warrant point files) will become available throughout the NBPD where it is accessible through the existing ESRI licensed software such as ArcMap and ArcCatalog.

Results

We tested the model by running a sample of 66 "scrubbed" warrant records.

The output resulted in:

• All non-local warrants (2) were successfully filtered out.
• 92.2% (59 out of 64) of the local warrants were successfully matched.
• 90.6% (58 out of 64) were matched by the more accurate address point locator.
• 1.5% (1 out of 64) was matched to the less accurate street locator.
• Unmatched addresses were determined to be the result of data entry errors in the original record.

Conclusion

Essentially the core of the geocoding tool has been built and with satisifactory results under our conditions. However, due to time constraints and unforseen issues with warrant data, the tool has not been tested in the server ArcSDE environment. The tool has to be configured to become web server capable. That is currently beyond the scope of this project.