Credit Scoring techniques revealed

never-in-doubt

PROLIANT wrote: »

I am playing around with some formulae then going to write a program in C#/Java just to see results, an avenue of computer programming I have never ventured in to - algorithms etc.
Data mining Star and Snowflake models is something I have worked with so I have a head start.

You clever sod! I actually understand what you're saying lol.... bear in mind most older score charts were created using C ++ anyway! :D You may actually sus it out mate, but only problem is getting it to match the formula of a lender - hmmmmmm (we need a mole). :p

C came out a year after I was born - NOT that old, really? !!!

soul619

ElkyElky wrote: »

Those documents though have given a good enough insight in to the whole process but we really need some underwriters who are brave enough to share the inner most secrets of their job.

and risk getting the sack..... norfolk n chance will any underwriter comment

PROLIANT

soul619 wrote: »

and risk getting the sack..... norfolk n chance will any underwriter comment

Hi thanks for the input, the goal of this thread is research and education not to reveal trade secrets or to encourage fraud. The black magic of underwriting and score cards is something that interests me and I am sure many others, please look beyond the sinister ulterior motives.

soul619

PROLIANT wrote: »

Hi thanks for the input, the goal of this thread is research and education not to reveal trade secrets or to encourage fraud. The black magic of underwriting and score cards is something that interests me and I am sure many others, please look beyond the sinister ulterior motives.

when you work with fraud ulterior motives is all you see!

PROLIANT

soul619 wrote: »

when you work with fraud ulterior motives is all you see!

Ok, I take it you work in the financial industry then?

Debt_Free_Chick

PROLIANT wrote: »

Hi thanks for the input, the goal of this thread is research and education not to reveal trade secrets or to encourage fraud. The black magic of underwriting and score cards is something that interests me and I am sure many others, please look beyond the sinister ulterior motives.

Agree with you - but if lenders were to reveal their risk model, fraudsters would have targets to aim at. Still, I think there's an opportunity for better understanding, even if we don't uncover all the subtle details of the risk models.

:T

izools

The information I would glean from this is that the "scoring" system used my most major lenders is simply a system of monitoring customers for behavioural aspects, account conduct, the system already knows their status in regard to time at address etc and most high street banks also keep an eye on the up to date credit file via Insignt and / or CAIS.

This information is constantly changing and being updated and fed back into the "neural net", which is in turn constantly updating the categories / scores it puts certain types of applicant in to.

I would suggest that it is reasonable, that in at least some, if not the majority of, scoring systems, no one other than the computer will know what the criteria are.

I would conclude that the criteria are not definite and / or static, and even if someone is able to monitor the way in which the system monitors existing customer' accounts and transposes this data into scoring techniques for new applicants, the information gleaned from said system would be hard to interpret, potentially change on a month to month basis, and be nigh on impossible to transpose into a method of cheating the system.

PROLIANT

Debt_Free_Chick wrote: »

Agree with you - but if lenders were to reveal their risk model, fraudsters would have targets to aim at. Still, I think there's an opportunity for better understanding, even if we don't uncover all the subtle details of the risk models.

:T

I am not asking for a live example from a major UK bank, just some insight to attempt to de-mystify the scoring techniques used e.g. how many searches are considered too many and in what time scale, why are searches viewed as a negative, why cant banks use unrecorded searches to give a credit decision and if successful then apply a foot print? How they justify a rejection based on certain data sets etc.

soul619

PROLIANT wrote: »

I am not asking for a live example from a major UK bank, just some insight to attempt to de-mystify the scoring techniques used e.g. how many searches are considered too many and in what time scale, why are searches viewed as a negative, why cant banks use unrecorded searches to give a credit decision and if successful then apply a foot print? How they justify a rejection based on certain data sets etc.

the answers to some of those questions would help aid potential fraud

never-in-doubt

soul619 wrote: »

when you work with fraud ulterior motives is all you see!

Not at all. I work within banking financial crime and wouldn't be big headed if I said I was bl00dy good at what I do and trust me - you must have an open mind to succeed - your attitude will get you nowhere, but then that's probably why I earn 5x what you do!

Bear in mind your audience before making meaningless or otherwise posts.... ;)

I found £100k last night, no ulterior motive at all. Had I went in with that attitude then you know what? I'd still be there now trawling data reports searching for staff fraud when you know what? It took 3 hours using data abnormality filters against the actual banks own systems.... i.e. the £100k was in "limbo" within their own systems - no fraud, simple accounting problem.

I think my point here is made.