ARC10 Geocoding w Dual Ranges zone erroneous

[ccain]

I am using ARC10 desktop with Service Pack 1 installed and am Working on the GISTutorial 1 Basic Workbook prior to teaching it. As part of the lesson a new address locator is made using Dual ranges comparing street address with zip codes. I have found that the address compare with a 95-100% match even though the zip codes are totally in error. As long as the street matches and a zip has five digits the match occurs.

I saw where this was an error NIM055232 supposedly fixed in Service Pack 1 but it is still not working correctly.

[BradNiemand]

NIM055232 fixes the opposite of the issue that you are having, no matches would occur if the locator is built without all of the zone fields mapped.

There is an issue where the score is higher than expected when a locator is built with not all of the zone fields mapped.  Essentially, the locator assumes that, in your case, the city and the state field have a score of 100 and the zip has a score of 0.  This increases the total score of the address and therefore can be a bit confusing.

There are a couple work-arounds that we can try and use if you would like.  The first would just be to increase the minimum match score to something like 96.  Another option would be for me to provide an updated style for you in which you could build your locator from.

Just let me know what works best for you and sorry for the inconvenience.

Brad

[ccain]

NIM055232 fixes the opposite of the issue that you are having, no matches would occur if the locator is built without all of the zone fields mapped.

There is an issue where the score is higher than expected when a locator is built with not all of the zone fields mapped.  Essentially, the locator assumes that, in your case, the city and the state field have a score of 100 and the zip has a score of 0.  This increases the total score of the address and therefore can be a bit confusing.

There are a couple work-arounds that we can try and use if you would like.  The first would just be to increase the minimum match score to something like 96.  Another option would be for me to provide an updated style for you in which you could build your locator from.

Just let me know what works best for you and sorry for the inconvenience.

Brad

I am not sure I exactly understand what you are saying. When I set up the Address locator my Reference Data feature [Pittsburgh streets] did not have a field for State and City and the Field Map defaults to <none>. If I understand you correctly you are saying that because the field map is <none> it takes this as a higher value or 100 match and ignores the zip. Thus as long as there is a 5 digit zip in the zip field it will match. [Those addresses with only four digits in the zip field failed]

Charlotte

[BradNiemand]

Yes you are correct.  The minimum match score out of the box is set to 85 so because of the issue of the unmapped fields, in your case city and state, you will get matches for 5 digit zip codes entered that are incorrect.

One work-around is setting the minimum match score of the locator to a higher value like 96.  The other would be to modify style to help alleviate the issue.

Brad

[ccain]

Okay, How do I modify the style? Create a new field in my reference data for State and city and set it to zero?

Charlotte

[BradNiemand]

I can update the style for you.  Would you like the score to be much lower or not return anything at all?

Brad

[ccain]

The score should be lowered.
In this case I am trying to teach how to find the errors and evaluate them plus how to use the geolocator. Although I think the way the zipcode is ignored is an error. You cannot assume that the address you are trying to locate is in the range of the geolocator. For instance if you were trying to geolocate participants at a local event, you could get participants from out of state thus with a valid zipcode that should not match to your locator.

[BradNiemand]

The score should be lowered.
In this case I am trying to teach how to find the errors and evaluate them plus how to use the geolocator. Although I think the way the zipcode is ignored is an error. You cannot assume that the address you are trying to locate is in the range of the geolocator. For instance if you were trying to geolocate participants at a local event, you could get participants from out of state thus with a valid zipcode that should not match to your locator.

A correction to your statement above.  The problem isn't that the zipcode is ignored, it is that the city and state are not ignored.  Let me clarify.  With zones (city, state, and zip), the locator tries first to search for the zip in the dictionary (think of the dictionary as exact match even though that is not exactly true).  If the zip is found, then it uses city and state ONLY for scoring, not for restricting search.  It then tries to find the street in this zipcode.  If the final score of the search is higher than the minimum match score, then it will stop and return the matched address.  If not, it will try to search for the city and state in the dictionary and use zipcode only for scoring.

So what happens is that the first search fails because there is no street in that zipcode that matches higher than the minimum match score.  Therefore the locator "falls back" to the city and state.  What happens is the locator compares "empty city" and "empty state" to empty dictionary and therefore matches.  It then reduces the score because the zipcode is wrong but the score is not reduced by very much because it thinks that the city and state are correct in this case and gives them both a score of 100.

I have attached a style that does not include city and state that will give you some better results for now.

[ccain]

Thank you Brad for the style and for the extra explanation. I now see why it is doing it this way. I appreciate it.

Charlotte Cain

[AndrewMaracini1]

I'm having an issue as well. I have a table with From To, Street Address, City, State, but no zip. When I run the built-in US Streets Geocoder I get a match of about 72%. When I try to create my own locator with dual ranges it fails completely with 0 matches. I've tried using different fields for my street address to no avail. I'm attaching a small subset of the data for your review.

thanks,

Andy