Skip to main content
. 2020 Nov 30;3:543405. doi: 10.3389/frai.2020.543405

FIGURE 2.

FIGURE 2

Average recall and precision for doctors and for the Babylon Triage and Diagnostic System (Babylon AI) for different threshold parameters, showing the operating point typically used (mode 1) and an alternative operating point (mode 2) that favors higher precision at the expense of reduced recall. Varying the internal thresholds allows the model to behave more similarly to different individual doctors, while maintaining a high level of performance, suggesting that it is not overly tuned to a particular operating point.