Performance in AGP in AML 7.1

[upadhi]

Hi all,

I am new to AML 7.1. Need a view on how we should map headers to scenarios.

Example:

i have 2 scenarios where I need to raise alert when a)party is older and large transaction amount b) party is minor and large transaction amount.

For these 2 scenarios, I have 2 approaches:

1. Approach-1:

one header: party_header (in here, the prep table will have all the transactions and in scenario code we are using age filter)

2. Approach-2:

2 headers: minor_party_header and older_party_header (in here, there will be 2 prep tables created on basis of age)

In which of the 2 approaches, i will get optimized performance??

[Patrick · Accepted solution]

I guess you will also have other TM scenarios and need anyway to create a prep file with all transactions. If so then I'd go for a single prep file - just make sure that your selection on age comes first in your scenario definition and I'd expect such a selection won't add much to runtime.

[upadhi]

Thanks Patrick. 

One more issue i am facing which is around process of transaction keys. For account level alerts, I am able to generate the transactions. however, it is not saving the transaction keys and because of the alert/entity level aggregates are not coming. could there be any specific reason i should look at?

[Patrick]

@upadhi 

The last hands-on AML implementation is for me now already too long ago to remember things on this level of detail.

I suggest that you create a new question for this so it gets attention. Because there aren't many people here in this forum answering SAS AML questions eventually also contact SAS TechSupport directly if you get stuck. 

[manishiiita]

Hi @upadhi, are you still stuck in this?

You can try following, try looking at replication days parameter. Ideally, it should be equal to maximum look back period. If it is set less than lookback period there will be problems in transactions replication.

Another reason could be, the triggering transactions are not being saved properly at scenario level. Please look at your scenarios code, compare with a OOB scenario. Another thing you can do is, try to access triggering transactions data set in data/stage/alerts folder. Check if the transactions are saved properly there.

Please let me know in case of any further help required.

[upadhi]

hi @manishiiita , Thanks for your reply. I was able to resolve this issue. Still thanks a lot!

[upadhi]

Hi  @manishiiita,

Thank you for your answer. This is resolved. However, there is another thing we are stuck on:

My AGP is scheduled to run Monday to Thursday in a week. However, if the AGP is running on Monday, it should generate alerts for transactions happened on last Friday, Saturday and Sunday. 

As per AML process, for daily scenarios, if AGP is running for Monday, it will generate alert only for transactions happened on Monday.
The requirement is that on Monday, it should generate alert for last weekend also.

Is there any optimal way to do this?

[Patrick]

@upadhi 

Just as an opinion: The optimal way is to run the daily AGP every day. If feels this requirement is mainly motivated by investigators trying to have a week-end but also to stay within their SLA. I'd consider it as a much better option to adopt the SLA for alerts raised on Friday, Saturday and Sunday than to try and tweak the AGP process.

I don't say it's impossible to setup the AGP accordingly but just imagine all the stuff you would need to do to run with a runasofdate for Monday but then on Mondays having the need to A) add more days worth of transactions and B) the need for extra logic in every single scenario (basically execute the scenario 4 times, once per day of transactions). ...or are you going to execute the AGP 4 times simulating past runasofdates so that suppression works like for other days. And then you've got also the challenge of DIM data changes. Are you for Friday transaction alerts going to use the data as per date of the transaction or as per runasofdate of Monday.