Overview

This article explains what the Prediction Results panel shows when a VRIFY Predict Experiment is added to a 3D Slide. It covers how to read prediction accuracy, depth accuracy, and feature importance, so you can confidently present and discuss DORA's outputs.

Note: This article applies to VRIFY Present Beta, which includes features not available in VRIFY Present Legacy.

What Is the Prediction Results Panel?

When you add a VRIFY Predict Experiment to a 3D Slide, the Prediction Results panel opens automatically. It summarises the key outputs of DORA's prediction model for that experiment, including how accurate the model is and which input features influenced the results most.

The panel contains four sections:

Target Minerals — the minerals the model was trained to predict
Prediction Accuracy — how reliably the model identified mineralised areas
Depth Accuracy — how well the model performed at depth
Feature Importance — which input features shaped the prediction most

Step-by-Step Instructions

Open the Prediction Results Panel

The Prediction Results panel opens automatically when an experiment is added. If it has been hidden, you can open it by:

Open the Predict section in the Object Settings panel.
Click the experiment name to open the Prediction Results panel.
Click the eye icon beside Prediction Results to show the panel.

Reading the Prediction Results Panel

Target Minerals

The top of the panel shows the minerals and grade thresholds the experiment was trained to predict — for example, Zn >= 182 or Cu >= 175. This tells you what the model was looking for when it generated the predictions.

Prediction Accuracy

Prediction Accuracy shows how often the model correctly identified both mineralised (positive) and unmineralized (negative) area, displayed as a percentage with a label:

Underfitted — the model may be too simple and could be missing patterns
Optimal — the model is performing well
Overfitted — the model may be too closely fitted to the training data and less reliable on new areas

For presentations, an Optimal score means the results are reliable and worth presenting with confidence. If the label shows Underfitted or Overfitted, consider flagging this with your geology team before presenting.

Learn more about overfitting and underfitting and prediction accuracy in the DORA article collection.

Depth Accuracy

Depth Accuracy shows how well the model predicts mineralisation below the surface, displayed as a percentage with the confidence label High, Medium, or Low. A higher score with a High label indicates the model's sub-surface predictions are reliable.

Learn more about depth accuracy in the DORA article collection.

Feature Importance

Feature Importance shows which input data, such as geochemistry, geophysics, or geological structures, had the most influence on the model's predictions, expressed as a percentage and ranked from most to least influential.

Each feature also includes a SHAP (SHapley Additive exPlanations) Distribution chart. This shows whether high or low values of that feature pushed predictions toward or away from mineralisation:

Data points to the right of centre → that feature increased the likelihood of mineralisation
Data points to the left of centre → that feature decreased the likelihood
Red/pink points → high values of that feature
Blue points → low values of that feature

For presentations, Feature Importance helps your audience understand what the model found most significant, and why certain areas scored higher than others.

Learn more about feature importance in the DORA article collection.