Building models with SAS Enterprise Miner, SAS Factory Miner, SAS Visual Data Mining and Machine Learning or just with programming

How about using probability instead of score?

Reply
New Contributor
Posts: 3

How about using probability instead of score?

Typically companies use score for credit approval, loan approval, and many other tasks. People often say approve them if score is above 443, or reject them if score is lower than 339, etc. However, scores provided by different vendors and different models have different meanings. A score of 600 from vendor A and from model B usually have different risks of default. I am thinking that if it is a better business practice to talk about probability, or likelihood of default. Tell decisions makers the likelihood of default is 50% by vendor A, 60% by vendor B, or 90% by model C. It will be easier for the decision maker to evaluate the risks, he can easily decide to chose the minimum, maximum, or average. It is also easier for developers to use some techniques to compare and combine risk assessments from different and ad hoc sources.If the company chooses to use score from another vendor or a new model, there will be no need to tell people that score criteria has changed, because the the criteria of likelihood doesn't change.

Valued Guide
Posts: 3,208

Re: How about using probability instead of score?

If you are referring to the fico score as numbers,  Credit score in the United States - Wikipedia, the free encyclopedia It is hiding the probability of likelihood of default.

An estimated as probability is a better figure. The difficulty is that it will  change on the involved amount and other data. That is left to the credit company. For a good comparison all those kind of predictors should be standardized and visualized. Not an easy one.

Getting to disconto and interest it would be better to use calculation base on seeing the time as continue not discrete. Even that simple better business pactice is hard to get accepted. As hard as statue or metric measurements. Liters or gallons, miles km of kn.

http://en.wikipedia.org/wiki/FICO#FICO_score

---->-- ja karman --<-----
Ask a Question
Discussion stats
  • 1 reply
  • 264 views
  • 0 likes
  • 2 in conversation