Published Multirole Locator results

1795
17
12-16-2019 03:00 PM
JoeBorgione
MVP Emeritus

ArcGIS Pro 2.4.2

I have a multi role locator with point address, street address-alt name and poi-altname roles.  The locator it self works great but I'm getting inconsistent returns when I add it as a published service.  When I use the multi role locator in the locate pane, I get one hit:

It's matched on the point address role, and that's just what I'm hoping to see.

When I add the published locator, sometimes \it gives two returns one on the Street Address, and a second on the Point address. To me, this is odd, especially when  the 'B' result is the address point:

Here I am back at my house, and I'm getting a single return, on the point address:

The inconsistency has me scratching my head.  Any ideas where things are getting goofy on me?

And if that's not weird enough, I have a list of 2500 addresses, and when I geocode against the locator, I get a 96% match rate.  When I match against the published service I match 0 (zero)....

Shana Britt

Eric Anderson

Brad Niemand

That should just about do it....
17 Replies
BradNiemand
Esri Regular Contributor

Joe,

Likely the limitations you are seeing are because of the older version of the software.  Please let me know once you have tried it against 10.7.1.

Brad

0 Kudos
JoeBorgione
MVP Emeritus

Probably after the new year.  Thanks for your help Brad!

That should just about do it....
0 Kudos
MichaelVolz
Esteemed Contributor

Is the Create Locator tool in Pro for multi-roles designed to just return 1 result which is the closest to the center of the map?  What if you have 3 identical addresses in a large jurisdiction where you would want the address locator to return all results for as you would zoom to each result individually?

0 Kudos
BradNiemand
Esri Regular Contributor

Michael,

If the addresses have the same attributes and are from the same role then all of them will be returned.  The first one will just be the closest to the center of the map.  Even if the addresses have the same attributes but are from differnt roles and are not spatially near each other, all the candidates will be returned.

The logic for deduplication of identical candidates happens when you have identical attributes across different roles where the locations are very close to each other.  Essentially we want to remove the StreetAddress match if we have a PointAddress match that represents the same address but with a more spatially precise location.

Brad

JoeBorgione
MVP Emeritus

Brad- is the 'center of the map' logic an ArcGIS Pro feature?  I ask because of this scenario:

Using the 'old style' create address locators, I have two address point locators: one is based on the General - Single Field style, and the other is based on US Address - Single House Subaddress style.  Both of these are members of a composite address locator.  

I just geocoded a list of addresses using the published composite service and as I'm reviewing the results, I notice an address that shows as a tied.  There are 6 points with the exact same address of 1860 E 9400 S (don' t ask me...)  In my screen cap below, you can see the 6 selected address points and the one highlighted is the one the geocoder picked as the match.  That looks to me to be the closest to the center of the map, at least the closest to the center of extent of my data which is where I was when I geocoded the addresses. Is this an example of what you mean by center of the map logic?  

Brad Niemand

That should just about do it....
0 Kudos
BradNiemand
Esri Regular Contributor

Joe,

The center of the map is only used when performing interactive geocoding, not batch geocoding.  What you are seeing is just a coincidence because when batch geocoding it will just choose the first candidate if there is a tie.

Brad

0 Kudos
JoeBorgione
MVP Emeritus

Interesting.  How is first candidate determined?  I would have thought the lowest value  Object ID would be used...(My definition of 'first')

That should just about do it....
0 Kudos
BradNiemand
Esri Regular Contributor

Joe,

There is not a really good answer for this.  It depends on a lot of factors but the best way to think of it is that it is random.

Brad

0 Kudos