Fig. 6: Application of the conformal deep learning model for managing heat-related risks. | Communications Engineering

Fig. 6: Application of the conformal deep learning model for managing heat-related risks.

From: Degrees of uncertainty: conformal deep learning for non-invasive core body temperature prediction in extreme environments

Fig. 6: Application of the conformal deep learning model for managing heat-related risks.

The plot shows how the model generates alerts by assessing the 68% prediction interval for core body temperature (CBT) across safety thresholds. Shaded areas represent the 68% confidence interval around each predicted CBT value, capturing the range within which the actual CBT is expected to fall. Alerts are triggered when prediction bounds exceed customizable safety thresholds, as shown in this example with unseen test subjects. Alert states are color-coded: Nominal (36.5 °C–37.75 °C, green), Advisory (37.75 °C–38 °C, yellow), Caution (38 °C–38.9 °C, orange), and Warning (above 38.9 °C, red). Confidence levels and thresholds are fully customizable, allowing the model to be adapted for various elevated heat-risk working environments where confidence-driven alerts support proactive safety interventions.

Back to article page