Model Connecting
Connecting to server...

Interactive simulator | Two-stage ML pipeline | XGBoost

Bandgap Prediction

Select a material from the test set. The dropdown reads target.json server-side, then the selected material id is matched against label.json and sent through the Model backend.

Model Connecting
Initializing the model health check
The runtime health probe is still pending.
Waiting for the model endpoint to report healthy status.
PLEASE WAIT

Threshold

The dissertation lowered the classifier threshold to 0.28 to protect nonmetal recall during screening.

Confidence

PI90 and PI95 represent split-conformal prediction intervals calibrated on held-out entries.

Risk tolerance

A practical acquisition score trades predicted Eg against interval width, mirroring uncertainty-aware screening.