Fuzzy match using a string variable between two large datasets

[art297]

Lan,

For three character names, I have found the compged function to be inefficient. However, SAS provides a number of alternatives (e.g., soundex).

However, there are additional steps you can take to limit probable noise before looking for matches (see, e.g.: http://ftp.sas.com/techsup/download/observations/obswww15/obswww15.pdf ).

Finally, it would definitely help if you were matching with a set of valid company names. It sounds like you are trying to clean up both data sets simultaneously which, methinks, would only add more confusion.

Art

[LanMin]

Thank you Art for sharing the resources and your advice ! I will read them carefully.

A little more details on my data:

main data: contains the universe of publicly traded U.S. firm names (I kept one unique record per firm, similar to your code).

second data: contains publicly traded company's customer name, these customers are themselves companies,

My main data has accounting data (e.g. firm cash holdings, debt, assets etc) of each firm, my goal is to get such info for my customer data, hence, I must match these two data sets.

In response to your latest comment, as far as I know both data contain valid company names, however, the database may not be build to users' satisfaction, the name abbreviation etc may not be used consistently across these two data sets.

Thanks again !

Lan