Best Practice of Ministry of Customs and Commerce Fraud - From Beginning to End SAS Approach

[ertr]

Hello everybody,

I want to make some research about Fraud in more specificly Customs and Commerce. Can somebody help me to understand the logic better about that subject? Any paper which you can suggest me or any apporaches which I can use it. What people usually do about this topic in the field and etc. I would be happy, if I can get some direction from you.

Thanks,

[Reeza]

Your question is too broad.

Are you using a SAS fraud analytics tool, or Base SAS? Are you interested in subject matter material - methodology of detecting fraud, or the actual mechanics of this, such as the code and detailed process?

[ertr]

Hello,

You can think that there are only Enterprise Guide and Enterprise Minerr products available. Do you have any suggestion about Fraud Analytical tools which makes this process better?

More specificly;

How can I create rules to catch Fraud better?

Are there any specific generic rules in this area?

Containers, duties, taxes, product weight, amount etc., which input can help me to create rules?

How can I perform anomaly testsin my data? What can I do except from Benford? -> https://support.sas.com/resources/papers/proceedings/proceedings/sugi27/p249-27.pdf

How can I interpret from Descriptive Analysis? Are there any specific report formats exist?

Can I benefit from Decision Tree in Enterprise Miner? (I am currently working on) Or what else can I benefit from Enterprise Miner? Logistic Regression?

Is it possible to create a Derived variable? 

In order to examine the data better, What is the preffered observed fraud rate? %2, % 5?

Can I use the analysis like PSI in this area? If so, how can I use it? -> http://support.sas.com/resources/papers/proceedings10/288-2010.pdf

@manishiiita ,

@Patrick

@SASKiwi 

@JuanS_OCS 

@Capt_VA_SAS 

@Ksharp

@PGStats 

Thanks

[JuanS_OCS]

Hi @ertr ,

very good question. True that it is broad but we can also tackle it in the same way.

In this sense, SAS offers several tools to work with Fraud. Not only SAS Base or SAS Miner.

For starters, in SAS 9.4, SAS offers a working framework namely SAS Fraud Framework, composed of Base, Miner, Forecast, DI, SNA (Social Network Analysis), FINCRN (Financial Crimes) and SAS Studio. Very powerwul. In the top of this framework, SAS or you can build the solution. A very popular one is Triage.

In addition, in SAS Viya, you have available as well SAS Visual Investigator, an advanced version of Social Network Analysis.

In any case, I would like to recommend you to contact your SAS Account Manager, so he/she can put you in contact with the expert sales/pre-sales and they can tell you more about what can be available, and better for your business requirements.

I hope I can help you with an initial direction.

Do you have any further questions?

Best regards,

Juan

[SASKiwi]

In my view, every company's fraud detection requirements are different and what SAS tools would be best for you depends on various use cases you have. I can't give a recommendation on a particular tool without knowing the specifics of each use case and trying to understand the unique signature of a particular type of fraud.

My background in fraud comes from the banking industry and especially credit cards. One particular project I was involved with was credit card "bin attack" fraud. This is where fraudsters use software freely available on the internet to generate card numbers and expiry dates automatically and then test these on vulnerable merchants' websites. They will rapidly test the card numbers in a certain range automatically with small test dollar amounts until they find one that works.

It turned out that Base SAS was the best way to tackle this type of fraud because it required processing huge card transaction logs and looking for card numbers repeated in the same range in a short space of time. Successful card transactions could then be tagged as suspicious if they had been preceded by many failures with almost the same card number and in the same time frame.

I give this as an example of the level of detail you need to go to before you decide what the best tool for a particular fraud use case.

[ertr]

Vaov, it is very nice to be appreciated about my question, thank you!

As far as I understand from your writing, we can strengthen the possibility of catching fraud with the following products?

• Forecast(Is there any specific product?)
• SNA(Social Network Analysis)
• FINCRN(Is there any specific product?)
• DI( I think DI is a product which can help us creating processes)
• SAS Visual Investigator

And Triage, can you give more detailed about Triage?

Thank you for your direction about Account Manager but is it possible to being a more specific?

Morever, how can we better capture fraud with Enterprise Miner and SAS Base(Enterprise Guide), except from specific products? I mean more purely?

What should we imply when we call Anamoly analysis?

• Benford
• Variable Clustering
• Correlation
• Rapid Predictive Modeler(RPM)
• Population Stability Index (PSI)
• K Nearest Neighbors
• Logistic Regression
• Decision Tree
• Support Vector Machine
• ETC.

Specific papers, codes, links or projects as above?

Yes, you helped  me a lot and motivated me ask some questions, thank you,

@WendyCzika 

@Jay54 

@Renato_sas 

@manishiiita

@SASKiwi 

[manishiiita]

@ertr , pardon for delayed response. I was too busy to look into SAS community in past couple of weeks.

Coming to your question, as others mentioned, your question is too broad and covers multiple domains. One thing I clearly get from your questions is that, you should focus on business understanding first. How the businesses work, how their data is important to a data scientist, what is the purpose of performing different operations on data etc. should be the main line of thought then only these technical things will make sense. 

I have mainly worked on AML & Fraud solutions. When I talk about a technical code / scenarios / rule for detection of Fraud / ML, it directly comes from the business requirements. So, first it's important to understand the business and possible fraud / ML cases. Then, we think about creating rules to identify / block such transactions. These business rules are then replicated into technical jargon. 

Your other questions are very specific to data analytics. I think you should try to post these questions separately, experts will help you getting an idea about it. Try to break your questions into small ones and post it on their respective forums.

About different products of SAS, yes there are many products of SAS available in market and being used by customers. some of them covering your question are SAS AML, SAS Fraud Management, SAS Risk Assessment, SAS SNA, SAS Visual Investigator etc. To gain good understanding of any solution / tool of SAS, it will take couple of months to years.