I trained a model that predicts customer's income given the features:

  • age, declared income
  • number of oustanding instalment, overdue total amount
  • active credit limit, total credit limit
  • total amount

The output is a prediction: lower-upper bound for a customer: e.g. [8756-9230]

Metrics used:

  • NIRDM - not in range distance mean - how far the value is from the closest bound (on average) for values out of range(similar to true negative)
  • in-interval - percent of tested values that actually happen to be within the range(similar true positive)

e.g. NIRDM = 1.37, on avg values deviate by 0.37, less is better.

my question: what other metrics can be used to estimate the accuracy of the model.

I dont have deeper expertise on this 'range estimation task', maybe confidence interval?


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.