DQ Match and SPEDIS approach

[angorwat]

Hi

I have two datasets - file 1 (50K records ) and file 2 (800K records) both have customer names. I matched these names using the DQMATCH function.

After the match  the output have the following variables - Name1, Name2, match_code_name, match_code_name1.  I have close to 5000  records in the output file.

Then, I used the SPEDIS function between the variables name1 and name2 and rank the output record based on the rank value. I.E., Lower rank higher probability of matches, ‘0’ rank is exact match.

I just want to know at what value of rank I need to stop as I can see some good match at the value 67. I want to extract the best match, don’t want to dump these 5000 records as an output records.

Is there any other approach I can use.

Please advise.

[ChrisNZ]

There is no best method for fuzzy matches. It all depends on your data, and on the number of matching errors you are prepared to tolerate.

Using several matching methods, and possibly several iterations, rather than just one, is a good idea.

Creating a "level of confidence in the match quality" rating field is also a good idea.

Also look at function COMPLEV (and COMPGED) which I found to be sometimes better for my use than SPEDIS.

[angorwat]

Thanks Chris