I am working on a machine learning project that I want to rank each customer and put on a scale smt like one of those


I've got almost 20k customer loan proposal data (customer information (~100 feature) + result of the proposal), each proposal end up with a decision (approve, reject, conditional approve) but no "customer rating" is labeled during the evaluation process so I don't know "goodness" of a customer

I thought about clustering customers but I can't be sure about number of clusters and I would really need of a expert who will put each cluster into related colors. Even after, I have no clue how to differentiate 2 customers which fall into the same color zone (green vs super green)

  • 1
    $\begingroup$ You need to define "goodness". Is it probability of repaying the loan from 0 (0% chance) to 1 (100% likely)? I get the impression that's basically what a credit score is, but on a different numeric scale. Once you have that definition agreed upon, you would use historical data on that customer to compute the % of loans (or amount of money?) they repaid, and then that could be the target for a regression problem. $\endgroup$ – CalZ Sep 23 at 13:51

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.