Matching closest house number with dual-range street data

[ShaharLevenson]

Hello all,

I have a dual-range street data.
I want to build an address locator with ArcGIS 10.2.1 that will be able to match the closest house number.

Take, for example, a case in which the left part of the street has the following segments: 2-12, 20-40, 50-80. The locator will match the house number 6 without a problem but I want to match also house number 16 that does not exists in the data. In this case the match will be to either house number 12 or 20 that are the closest house numbers in the data.
I can partially do it by lowering the spelling sensitivity but than I will have a problem with the street name.
Is there something that I can change in the locator style in order to accomplish what I need?
Also, I would like to maintain parity (matching odd numbers to odd-numbers street side) it that's possible.

Thanks,
Shahar.

[JoeBorgione]

Hello all,

I have a dual-range street data.
I want to build an address locator with ArcGIS 10.2.1 that will be able to match the closest house number.

Take, for example, a case in which the left part of the street has the following segments: 2-12, 20-40, 50-80. The locator will match the house number 6 without a problem but I want to match also house number 16 that does not exists in the data. In this case the match will be to either house number 12 or 20 that are the closest house numbers in the data.
I can partially do it by lowering the spelling sensitivity but than I will have a problem with the street name.
Is there something that I can change in the locator style in order to accomplish what I need?
Also, I would like to maintain parity (matching odd numbers to odd-numbers street side) it that's possible.

Thanks,
Shahar.

I guess you can do one of two things:

1.  Improve your street data to better match the addresses that are along the segments
2.  Try turning down your candidate scores.  That might get you a partial hit on house 16...

[KimOllivier]

I managed this problem by dissolving all the roads into single streets and added the min and max, odd and even calculated for the whole street to make a second fallback locator that would have all the gaps covered by the range. It might be a bit more approximate but you do get a match. You could also have some defaults based on the length of the street and the average house frontage width.

If you wanted to refine it, you could create a route system, calibrate with the existing address ranges and then use the measures to predict the start and ends on each segment to repopulate the original segments.

[ShaharLevenson]

These are good suggestions.
I indeed tried lowering the candidate score and it gave me exactly what I wanted. It even keeps parity!

I think that the dissolve method wouldn't quite work for me since I also need to match house numbers that are completely out of street range (eg., matching house number 2030 to a street that has only a range of 2-10). Obviously, the geocoding results aren't accurate at all, especially since the out-of-range house number probably comes from a typo, but that's another story...

Thank you kimo and jborgion!