+ Reply to Thread
Results 1 to 6 of 6

Thread: Why the difference: 10 vs. 9: 100% sensitivity, 100% min match, diff results

  1. #1
    Harry Bowman
    Join Date
    Apr 2012
    Posts
    142
    Points
    1
    Answers Provided
    1


    0

    Default Why the difference: 10 vs. 9: 100% sensitivity, 100% min match, diff results

    10: US Dual Ranges with ZIP and Alt Name, 100% spelling senstitivity, 100% minimum match score
    97006 match
    380 tie
    22073 unmatch

    9: US Streets with Zone and Altname, used ZIP as zone, 100% spelling sensitivity, 100% minimum match score
    96211 match
    3950 tie
    19298 unmatch

    With the same fields mapped and 100% senstivity and 100% minimum match, I'd expect the same results. In what way is the ver. 10 locator more "picky" under those circumstances? I haven't gone diggiing yet and really don't want to. My employer refuses to move to 10 without understanding the difference, however.

  2. #2
    Brad Niemand

    Join Date
    Nov 2009
    Posts
    251
    Points
    37
    Answers Provided
    4


    0

    Default Re: Why the difference: 10 vs. 9: 100% sensitivity, 100% min match, diff results

    The locator properties are set to optimal values out of the box. Try comparing the 9.3 locators to the 10.0 locators with the default values for minimumMatchScore and SpellingSensitivity. See what you get with this.

    Brad

  3. #3
    Harry Bowman
    Join Date
    Apr 2012
    Posts
    142
    Points
    1
    Answers Provided
    1


    0

    Default Re: Why the difference: 10 vs. 9: 100% sensitivity, 100% min match, diff results

    Quote Originally Posted by brad5993 View Post
    The locator properties are set to optimal values out of the box. Try comparing the 9.3 locators to the 10.0 locators with the default values for minimumMatchScore and SpellingSensitivity. See what you get with this.

    Brad
    I'm not allowed to do this. HQ demands 100% match for reference data testing. They also demand manual editing of the input data to match the substitutions the geocoder will do automatically. They were burned in the past with 80% min match score geocoding addresses into the wrong state based on street address + ZIP so do not want us to use fuzzy matching, ever.

  4. #4
    Brad Niemand

    Join Date
    Nov 2009
    Posts
    251
    Points
    37
    Answers Provided
    4


    0

    Default Re: Why the difference: 10 vs. 9: 100% sensitivity, 100% min match, diff results

    Have you tried the same test without the alt name table for both? I would be curious to see the results for that.

    Brad

  5. #5
    Harry Bowman
    Join Date
    Apr 2012
    Posts
    142
    Points
    1
    Answers Provided
    1


    0

    Default Re: Why the difference: 10 vs. 9: 100% sensitivity, 100% min match, diff results

    Quote Originally Posted by brad5993 View Post
    Have you tried the same test without the alt name table for both? I would be curious to see the results for that.

    Brad
    No I haven't. I can try that when I have some time, maybe soon. I ran the geocoding on two sets of street data. In one set, the data was all one feature class, with alternate names as alternate features with identical geometry. In the other, the data had an alternate name table. When I get a chance, I will look at the results again. I do know that some things that looked perfect were scored at something like 99.4% when I cheated and did fuzzy as an experiment.

  6. #6
    Brad Niemand

    Join Date
    Nov 2009
    Posts
    251
    Points
    37
    Answers Provided
    4


    0

    Default Re: Why the difference: 10 vs. 9: 100% sensitivity, 100% min match, diff results

    We do apply a small score deduction for spelling mistakes which is what you are seeing with a score that is above 99.

    Brad

+ Reply to Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts