Fuzzy Match for Names

Question

Hi

I have problem with the fuzzy match tools. I'm trying to use fuzzy match for names, examples as below:

ABASMAYATI ABDULAZIMASMAYATI BINTI ABDULAZIMAZIZANG BIN IBRAHIMAZIZANG IBRAHIMCANDAY LOH FOONG MANGLOH FOONG MANG

However current settings, pick up wrongly for example like below:

Current setting Match threshold 90% with match style: Name

CDCHIAH YUN CHINGCHIA SHYAN CHINGLOKE HENG FATTYLEE CHONG FATTYNURA ALIYANA BINTI MOHAMMADADNURA LIYANA BINTI MUHAMADADWUN PEI KENGWUN PEI KANG

Is there a way to fix this?

apathetichell · Answer

have you considered parsing out the  name into components and then trying a fuzzy match? I think the way it searches it gives equal weight to variation at different point in the name - however, as you've pointed out differing sections of the name may be more important. In your case, you'd probably want a 100% match on "last name" and differing levels of fuzzy match on the other parts.

ArtApa · Answer

Hi @alkafalhas - I don't believe this is achievable on the level that you described. The problem is that something that you describe as a "wrong" match is objectively a better match. I tried different Match Functions and "wrongs" always tend to have a better matching score:

I'd recommend to look at other features in your data set or on the process holistically to achieve a desired outcome. Less elegant idea would be to create a table of exceptions and use it to remove "wrongs" from the workflow before Fuzzy Matching.

Maskell_Rascal · Answer

Hi @alkafalhas

Can you provide either your workflow or sample data?