Open your black box classifier

The transparency of machine learning models is central to good practice when they are applied in high stakes applications. Recent developments make this feasible for tabular data, which is prevalent in risk modelling and computer-based decision support across multiple domains including healthcare. Important motivating factors for interpretability are outlined and practical approaches are summarised, signposting the main methods available, with pointers to the supporting literature. A key finding is that any black box classifier making probabilistic predictions of class membership from data in tabular form can be represented with a globally interpretable model without loss of performance.