Author Topic: Filtering vs Credit Model Algorithm  (Read 3034 times)

megamx26

  • Newbie
  • *
  • Posts: 28
    • View Profile
Filtering vs Credit Model Algorithm
« on: August 07, 2015, 11:08:13 AM »
I'm curious what the differences are and if there are any examples of how a credit model algorithm works.

Rob L

  • Hero Member
  • *****
  • Posts: 2043
    • View Profile
Re: Filtering vs Credit Model Algorithm
« Reply #1 on: August 07, 2015, 08:10:48 PM »
FWIW I consider this a pretty good book on the subject:

Credit Scoring, Response Modeling, and Insurance Rating: A Practical Guide ...
By Steven Finlay

Then again I'm no expert.

lascott

  • Hero Member
  • *****
  • Posts: 1423
    • View Profile
    • Appreciate my post and want to try LendingRobot? URL below
Re: Filtering vs Credit Model Algorithm
« Reply #2 on: August 07, 2015, 09:21:33 PM »
I'm curious what the differences are and if there are any examples of how a credit model algorithm works.
I suspect this has been talked about various times in the past but I have not been around here very long.

It seems that filtering is pretty straightforward to understand but can be so restrictive it limits the number of notes you can get.  Everything is AND'd together and all must be true.  Consider "Public Records" ... if you filter for the ideal of 0 then you may miss notes that have 1 but it happened 60 months ago (5 yrs). Is that length of time enough to show whatever lead them to that is no long an issue?

I have never read any books on credit scoring but simplistically I believe it assigns points/weights to each credit attribute you care about (Public Records, Employment Len, Months since Delq, Nbr Accounts open, Credit Utilz %, % of Bankcards >75% of Limit, Earliest Credit Line (Credit Age), etc).   Bad example below but gets basic point across.
PubRecs = 0 then 75
PubRecs 1-5 then 35
PubRecs 6-11 then 25
PubRecs 12-23 then 15
PubRecs 24+ then 0

Then you add up the values of all the points/weights to see if it passes a threshold you set.  Or perhaps multiple thresholds. 

If Total < 450, then do not invest
If Total between 450 and 475, then invest $25
If Total between 475 and 500, then invest $50
If Total > 500, then invest $75 

Credit scoring will let you find more notes that meet your guidelines.

(Variation to point system above could be opposition so the smaller the number the better).
« Last Edit: August 07, 2015, 09:34:56 PM by lascott »
Tools I use: (main) BlueVestment: https://www.bluevestment.com/app/pricing + https://www.interestradar.com/ , (others) Lending Robot referral link: https://www.lendingrobot.com/ref/scott473/  & Peercube referral code: DFVA9Y

AnilG

  • Hero Member
  • *****
  • Posts: 1090
    • View Profile
    • PeerCube
Re: Filtering vs Credit Model Algorithm
« Reply #3 on: August 07, 2015, 10:30:04 PM »
Steven Finlay book is quite good. In addition, if credit modeling and scoring is your interest, you might want to also consider reading:

Naeem Siddiqi, Credit Risk Scorecards Developing and Implementing Intelligent Credit Scoring, 2005
Raymond Anderson, The Credit Scoring Toolkit Theory and Practice for Retail Credit Risk Management and Decision Automation, 2007

Also, you can find lot of research papers on this topic using Google Scholar and SSRN.

FWIW I consider this a pretty good book on the subject:

Credit Scoring, Response Modeling, and Insurance Rating: A Practical Guide ...
By Steven Finlay

Then again I'm no expert.
---
Anil Gupta
PeerCube Thoughts blog https://www.peercube.com/blog
PeerCube https://www.peercube.com

rawraw

  • Hero Member
  • *****
  • Posts: 2759
    • View Profile
Re: Filtering vs Credit Model Algorithm
« Reply #4 on: August 08, 2015, 05:32:33 AM »
Do you remember that Y = MX + B formula from algebra?  The way I think of it high level is filtering is a bunch of "AND" operators, while the other method calculates a score based on variable and their coefficients.  It is much less blunt than a bunch of "AND" operators, which may eliminate a great loan because one variable outside your parameters is compensated for by exceptionally good elsewhere (perhaps their DTI is 25 vs your filter of 20, but they make $1 million a year in income).  This ultimately increases the volume that passes the cut.

Fred

  • Hero Member
  • *****
  • Posts: 1421
    • View Profile
Re: Filtering vs Credit Model Algorithm
« Reply #5 on: August 10, 2015, 02:03:31 AM »
Then again I'm no expert.

In academic and scientific community, "experts" show their research, findings, and theories to the world.

Unfortunately, in financial community, "experts" keep their lips tight.