how to match company name based on the first character of company name

[France]

dear all,

how to merge two tables based on the first character of the 'company name' variables.

bu using the 2 below tables as an example,

table 1

Type	NAME	COMPANY NAME	COMPANY NAME_1	COMPANY NAME - SHORT	PREVIOUS NAME
917574	ABBOT GROUP DEAD - 07/03/08	ABBOT GROUP PLC	ABBOT GROUP PLC.	ABBOT GROUP PLC	UNIGROUP
991144	ABBOTT LABS.GBP (LON) DEAD - DEAD	ABBOTT LABORATORIES	ABBOTT LABORATORIES	ABBOTT LABORATORIES	ABOTT LABS.GBP (LON)
671863	ABBOTT LABS.USD (LON) DEAD - 21/10/16	ABBOTT LABORATORIES	ABBOTT LABORATORIES	ABBOTT LABORATORIES	ABBOTT LABS. (XSQ)
926020	ABERCOM GROUP (LON)	ABERCOM GROUP LTD.	ABERCOM GROUP	ABERCOM GROUP LTD
911147	ABERDEEN STEAK HOUSES DEAD - 03/01/03		ABDN.STEAK HSES.GP.PLC.
902737	ABERDEEN TRUST		ABERDEEN ASSET MAN.PLC.
507161	ABDN.TST.WTS.A EXPIRY 30/09/99		ABERDEEN ASSET MAN.PLC.
961337	ABERFORTH SMALLER COS.C		ABERFORTH SMCOS.IT.PLC.
928988	ABERFORTH SMCOS.C		ABERFORTH SMCOS.IT.PLC.
960130	ABERFORTH SPL.C SHARE DELISTED 27/07/94		ABERFORTH
15295C	ABERTIS (IRS)	ABERTIS INFRAESTRUCTURAS	ABERTIS INFSTS.SA	ABERTIS INFRAEST	ACESA INTRAESTRUCTU(IRS)
29253E	ABERTIS (IRS) BONUS RIGHTS 03/08/04	ABERTIS INFRAESTRUCTURAS	ABERTIS INFSTS.SA	ABERTIS INFRAEST	ACESA INTRAESTRUCTU(IRS)
256273	ABGENIX (IRS) DEAD - 12/11/07	ABGENIX, INC.	ABGENIX INCO.	ABGENIX, INC.

table 2

code	psn_name
1	ABBOTT CCCC
2	ABERDEEN ww
3	ACESA des
4	APPLE
5	ABERFORTH ftrr
6	ABDN dggg

logic:

1splite each observation based on the 'blank' ',' '.' etc. in both tables.

2merge 'psn_name' in table 2 with each company name variable in table 1(i.e., NAME,  COMPANY_NAME, COMPANY_NAME_1,COMPANY_NAME_SHORT,PREVIOUS_NAME) if 1)both of the characters is the first character of the 'company name' variable,

and 2) both of the characters are same.

if the psn_name can be matched with any of the 'company name' variables, it can be recorded with the entire row.

3 record the matched table,  non_matched table 1 and non_matched table 2 separately.

so, for the 

code	psn_name
1	ABBOTT CCCC

the first character of psn_name variable is 'ABBOTT'. 

so it can match with 

991144	ABBOTT LABS.GBP (LON) DEAD - DEAD	ABBOTT LABORATORIES	ABBOTT LABORATORIES	ABBOTT LABORATORIES	ABOTT LABS.GBP (LON)
671863	ABBOTT LABS.USD (LON) DEAD - 21/10/16	ABBOTT LABORATORIES	ABBOTT LABORATORIES	ABBOTT LABORATORIES	ABBOTT LABS. (XSQ)

as the first character of the variable is 'ABBOTT' as well

the second example,

6

ABDN dggg

so it can match with 

911147

ABERDEEN STEAK HOUSES DEAD - 03/01/03

ABDN.STEAK HSES.GP.PLC.

 as the 'ABDN' is as same as the 'ABDN' in 'ABDN.STEAK HSES.GP.PLC.'

Could you please give me some suggestions about this? the sample has been added to the attachment.

thanks in advance.

[Reeza]

See FriedEggs solution here. I’ve used that a few times, usually get a 70% match rate.
https://communities.sas.com/t5/SAS-Procedures/Name-matching/td-p/82780

[France]

thanks Reeza, I will try it later(as my SAS is running the other program) besides, could you give me some suggestion if I really expect to match company name based on the first character of the company name? As my supervisor is asking me to match part of data handly. and I hope to improve the process by matching the first character through code

[Reeza]

It'll reduce the number of errors. I think I listed out the steps previously you need to do when doing a comparison like this:

1. Remove all exact matches

2. Do a fuzzy match - remove those

3. etc......

If you're at a University though you may qualify for a cheaper license or educational license for SAS Data Management Studio or Dataflux which can do the matching in a more robust manner. I didn't have great luck, but it depends on your data.

[France]

Dear Reeza,

thanks for your message.

is it the advice? 

This is a non-trivial problem because you end up doing N*M calculations which can be quite big. 

Some tips:

1. Upcase everything

2. Change all & or other symbols such as @ from text

3. Ensure Co and/or LTD are standardized, i.e. LTD vs LTD. vs Limited vs Limited

4. Do an exact match and remove those records

5. Remove all spaces and try a second round of exact matches

6. Do fuzzy matching using COMPGED/COMPLEV/SOUNDS LIKE/SPEDIS and equality of first letter

7. Do fuzzy matching using COMPGED/COMPLEV/SOUNDS LIKE

COMPGED works best in my experience

https://communities.sas.com/t5/General-SAS-Programming/Matching-Company-Names/td-p/439674

however, I confuse about several steps.

1.'Change all & or other symbols such as @ from text'  how to change the symbols? should I remove all the symbols? or replace them with the same symbol? 

2.'Ensure Co and/or LTD are standardized, i.e. LTD vs LTD. vs Limited vs Limited' how to standardize these words? Is there any code to do this step? particularly, will you remove the 'Co and/or LTD' words in a dataset, if few 'Co and/or LTD' words are recorded in the database?

[ballardw]

---

@France wrote:

dear all,

how to merge two tables based on the first character of the 'company name' variables.

bu using the 2 below tables as an example,

as the first character of the variable is 'ABBOTT' as well

the second example,

6

ABDN dggg

so it can match with 

911147

ABERDEEN STEAK HOUSES DEAD - 03/01/03

ABDN.STEAK HSES.GP.PLC.

 as the 'ABDN' is as same as the 'ABDN' in 'ABDN.STEAK HSES.GP.PLC.'

Could you please give me some suggestions about this? the sample has been added to the attachment.

thanks in advance.

---

You mean first word not character.

SCAN function will identify the first word of a string.

Do you have UNIGROUP in table 2? If so what would the output look like?

Is the program supposed to work for "ABOTT LABS.GBP (LON)" when it appears the match you are requesting would not work because ABOTT does not match the obvious ABBOTT in table 2?

Is the a preferred order of comparison between the 5 different name variables?

[Oligolas]

Hi,

you could use the LIKE SQL Operator but if you're in a professional environment or what you do can have a financial impact, I would rather recommend you to select distinctively the company names from table1 and 2, match them by hand in a separate dataset and afterwards to link all 3 tables together. From my point of view it would be much faster, reliable and less challenging than trying to identify afterwards which matches were inaccurate and why.

[France]

Dear ballardw,

thank you for your advice. what is the meaning of 'UNIGROUP' could you please explain it for me?