One comment

1. The problem of “black box” decisioning has reduced with Models allowing for clear outputs on the logic which weighted the decision in why it flagged a customer or account. While this can take some intepretation, it is possible to produce output in a relatively clear manner to inform an investigator on why a customer is potentially unusual. Similarly, with alert closure or qualification, a well detailed and documented output message can provide a clear auditable justification. The use of Machine Learning must be executed in a well controlled governance structure with control groups and 2LOD oversight. Rule based modelling will slowly then suddenly be replaced in the future.

   Log in to Reply