How I can Improve a Fuzzy match Score

[ByranCarter12]

Hello, 

I am trying to compare adresses each other with using Fuzzy Match Tool. 

you can see the result below :

Adress	Adress_matching	Id	MatchScore	MatchScore_Adress
605 3rd Avenue	605 Third Avenue	2	0	0
2203 Rowan Street	2203 Rowen St	3	85	85
1550 W Isaac DR	1550 West Isaac DR	4	91	91
340 S Patterson DR	340 South Patterson DR	5	84	84
1555 W 3rd ST	1555 Third Street	6	0	0
2310 S Winslow CT	2310 Sth Winslow CT	7	96	96
1110 S Covenanter DR	1110 Sth Covenanter DR	8	89	89
2430 S Brown AVE	2430 Sth Brown Avenue	9	91	91
1918 W 3rd ST	1918  Third Street	10	0	0

So, When the adress is written like 'Third' instead 3rd, the result return the value 0, even if the two adress are similars.

I've used this paramaters to do this fuzzy match 

So wich options can improve the result when there is a number written in word (1th ==>first , 3rd ==> Third ....). 

I've also an another question about the Fuzzy Match Tool, how i can avoid to having duplicates in the result, i give you this example :

Id	Id2	MatchScore	MatchScore_Adress
723	723	100	100
723	953	82	82
723	953	82	82
953	953	100	100
691	691	91	91
691	961	72	72
691	961	76	76

The Tool calculate the score with others Id and I don't know why. 

Thank you for your help.

[ArtApa]

Hi @ByranCarter12 - Fuzzy matching is an art. It is a very iterative process. Here are my comments:

1. You may Need to prepare your data set first. In your particular scenario I converted "3rd" to "Third".
2. Match everything that can be matched before Fuzzy matching
3. Start with the default Match Style (Address)
4. The longer the Max Key Length the lower the match score. For "Address" the default length is 8. You may want to reduce it to 6 or 4 to improve the score.
5. The Unique tool will remove duplicates.

Here is how your solution may look like:

You may want to watch the following videos to learn more about Fuzzy Matching in Alteryx:

• https://community.alteryx.com/t5/Videos/Fuzzy-Matching-Intermediate-Users/td-p/43852
• https://community.alteryx.com/t5/Videos/Fuzzy-Matching-for-Beginners/td-p/330575
• https://community.alteryx.com/t5/Videos/Introduction-to-Fuzzy-Matching/td-p/43884 

Enjoy!

[ByranCarter12]

@ArtApa 

Thank you, I appreciate your help.

I also had this idea to replace some abbreviations to improve the score. But when it comes to words like (third, first, fifth...), it might be long and boring to list all the possibilities to replace. There is no way in the options to ignore ordinal values.

In addition, the result obtained is 0 while there are similarities between the two addresses.

[ArtApa]

Hi @ByranCarter12 - Well, if you are to solve a business problem then nothing is boring. Building a lookup table is a valid approach. See example here: https://community.alteryx.com/t5/Weekly-Challenge/Challenge-235-Spell-It-Out-For-Me-Part-2/td-p/6409...

The way the Address Match Style works it ignores: Ave, St, Rd, Blvd etc. If you want to change the style to Name, then the Score will jump up for this particular record, but you may have a lower case somewhere else:

That's why I wrote, it's an art. This is why it's called Fuzzy.