Uday,
From my understanding, those APIs should work. Since the PME (probabilistic matching engine) was lifted from the MDM SE engine and embedded into the MDM Operational Server, the algorithm and it's interactions are somewhat consistent between the versions. So, if it's the Probabilistic search, it leverages a similar set of underlying matching and linking processes.
My 14 years of experience is primarily on the MDM SE side, but the documentation references the following 2 parameters of the API calls, which should enable you to get back what you want:
MinSuspectTypeThe minimum suspect type threshold for which matched records will be returned.
This value can be provided to override the default suspect type of a B match (a possible match). Valid values for MinSuspectType are defined in the CDSUSPECTTP code table:
- A value of
1
indicates a probabilistic match threshold type of Match
. <<< THIS IS THE ONE YOU WOULD WANT TO USE
- A value of
3
indicates a probabilistic match threshold type of Possible Match
.
- A value of
4
indicates a probabilistic match threshold type of No Match
.
MinScoreThe minimum score for which matched parties will be returned. This numeric value can be provided to override the default setting of a possible match (a B match). If the MinSuspectType is also provided in the request, then the transaction will use the MinSuspectType value instead of the MinScore value.
The
PartyID should be included in the result set from this kind of a search. If not, you may have to run an secondary call to get the party information.
If you look at the
ProbabilisticPersonSearchBObjType documentation (here), it mentions that it includes the TCRMPersonSearchResultBobj is returned. The documentation for TCRMPersonSearchResultBoby (here) shows PartyID as one of the response outputs. So, you should be good using that.
Obviously, if you're using ORG instead of PERSON the process is pretty much the same. PartyID is also a part of the TCRMOrganizationSearchResultBoby.
------------------------------
Tyson Carter
------------------------------
Original Message:
Sent: Wed November 04, 2020 11:34 PM
From: Uday Shankar
Subject: External Match in IBM MDM
Thanks Tyson for the quick response.
Could you please help me on IBM InfoSphere MDM AE (Physical MDM) as we are using AE.
SearchOrganizationProbabilistic
SearchPersonProbabilistic
As per my understanding these above could do the same. Please suggest.
Thanks
------------------------------
Uday Shankar
Original Message:
Sent: Wed November 04, 2020 12:25 PM
From: Tyson Carter
Subject: External Match in IBM MDM
If using IBM InfoSphere MDM SE (Virtual MDM) the best approach would be to leverage the Member Search (IxnMemSearch) with the parameters that you want to return the results "ASENTITY" with a minScore that is the same as your AutoLink Threshold.
Since the Member Search is Source Agnostic, the API would allow you to send the attributes for each record as Search Criteria. The key is that you would only really need to send the attributes referenced by the Algorithm for Comparison. By setting the "minScore" to your AutoLink threshold, you would end up only getting back results that would score high enough to "Link" to one (or more) of your Entities. Additionally, you could leverage the maxRows = 1 parameter to limit your results to just the single "best" match in the event that a record might partially match more than one Entity.
Key parameters to make this work:
segcodefilter = MEMHEAD, ENTLINK
gettype = ASENTITY
searchtype = ASMEMBER
minscore = [your AutoLink threshold]
maxrows = 1
Here's more information on the MEMBER Search API:
https://www.ibm.com/support/knowledgecenter/SSWSR9_11.6.0/com.ibm.mdshs.sdkref.doc/topics/r_sdkref_interactions_membersearch.html
------------------------------
Tyson Carter
Original Message:
Sent: Wed November 04, 2020 12:12 AM
From: Uday Shankar
Subject: External Match in IBM MDM
Hi All,
In Informatica MDM we have a feature called "External Match" that allows one to run source data through the matching process and obtain a golden record ID without actually loading the source data into the Hub.
Does IBM MDM have this feature?
Thanks
------------------------------
Uday Shankar
------------------------------
#MasterDataManagement