Credit risk teams are required to use models a human can genuinely follow, but the standard interpretable models used to meet that requirement, a basic logistic regression, a shallow decision tree, sacrifice real predictive accuracy to get there. Standard tools settle for a locally optimal set of features, leaving accuracy on the table that a black-box model would capture and a regulator won't accept.
iQ ML closes that gap: interpretable models built through systematic, optimal feature selection, reaching accuracy levels usually reserved for black-box approaches. One of these approaches was developed through a research collaboration with the Bank of Spain. Built for credit risk teams at financial institutions and the consultancies that support them.
