Katkat
Back

Determine Posterior Probability of Bad User Prediction

MetaData Scientist · Onsite
EasyData ScientistOnsiteStatistics & Math

Last updated

Meta Bayes theorem question on a bad-actor classifier, covering prevalence, sensitivity, specificity, false positives, positive predictive value, and the base-rate effect.

Scenario

Evaluating a classifier that flags bad actors

Question

5 % of users are actually bad. The model labels a user correctly with 95 % accuracy for both classes. If the model predicts a user is bad, what is the posterior probability the user is truly bad?

Hints

Direct application of Bayes’ rule with symmetric 95 % true-positive and false-positive complements.