What is a white-box model?

For a model to be considered white-box its features must be visible and the model build must be clearly explained. Decision-tree models are good examples.<br/><br/>

Black-box models, such as neural networks, boosting and random forest models are more complex and difficult to conceptualise. Although usually more accurate than white-box, they are more difficult to govern and any failings in the model become proportionately more difficult to justify.<br/><br/>

Companies need to find a fine balance between transparency, governance and accuracy.