Overview

Learning data defines which elements you are targeting and separates sample points into positive (mineralized) and negative (unmineralized) examples. These labeled points train the model to recognize spatial patterns and generate predictions.

In this step, you will select your target element(s) and review the recommended threshold used to classify your learning points.

What’s Improved

This step now includes an automatic, data-driven mineralization threshold recommendation.

DORA analyzes your learning data to suggest a meaningful starting cutoff, reducing guesswork during experiment setup.

What is Learning Data?

Learning data comes from the Learning Points Shapefile uploaded as part of your project data. This file contains sample locations with attributes such as element grades and elevation values.

At a minimum, each learning point must include:

Coordinates (easting and northing)
Elevation
Assay data for one or more elements

Most learning points are sourced from drill assays (de-surveyed into 3D space) or in-situ surface rock assays. Accurate spatial positioning is essential for reliable predictions.

DORA classifies each point as positive (mineralized) or negative (unmineralized) based on the selected threshold.

How DORA Recommends a Threshold

When you select a target element, DORA analyzes both the distribution of grades and how those grades are spatially arranged across your AOI.

Rather than relying on grade values alone, DORA evaluates how samples relate to one another in space. It identifies statistically significant clusters of elevated grades and distinguishes these from isolated high-value outliers.

The recommended threshold is designed to capture spatially coherent mineralization zones, not just the highest individual samples. This creates a more geologically meaningful separation between mineralized and unmineralized examples.

This threshold is a defensible, data-driven starting point to pre-filter inputs. However, it should be reviewed and validated against your geological understanding and exploration goals before finalizing.

Why This Step Matters

Geoscience Perspective

Separating mineralized from unmineralized data mirrors the way geoscientists interpret anomalies. It ensures the model is trained on examples that reflect real-world exploration targets, leading to more reliable predictions.

AI Perspective

Clear labeling allows the model to learn the spatial patterns that distinguish signal from background. A two-class structure provides a stable and interpretable foundation for generating predictions.

💡 Learning Data is the foundation for the Performance Breakdown, one of three output graphs generated by DORA alongside the VRIFY Prospectivity Score (VPS).

The Performance Breakdown evaluates how well your labeled examples helped the model distinguish between mineralized and unmineralized zones.

Step by Step Instructions

Open Set Up Learning Points
- From the experiment setup panel, click Step 2: Set up Learning Points.
Select Learning Points File
- Select the Learning Points file from the dropdown. If only one file exists, it is used automatically and this option will not be available.
  - (Optional) To upload a new Learning Points file within the DORA interface, click Manage Learning Points at the bottom of the dropdown. Note that it can take up to 15 minutes to process before it appears in the list. Learn how to format, upload, and manage your Learning Points file.
Select Elevation Field
- Choose the Elevation Field that matches the Z-coordinate column in your shapefile. This is used for 3D prediction.
Select Target Element(s)
- Select a target element from the dropdown. From here, DORA automatically generates a recommended threshold for classifying positive learning points.
- The histogram displays grade distribution:
  - The X-axis represents the minimum and maximum grade values.
  - The Y-axis shows the number of sample points within each grade range.
- You can move forward with the suggested threshold, or use the slider to adjust.
- (Optional) Add More Target Elements
  - Click Add Element to include additional elements and define individual thresholds.
  - If using multiple elements, choose how they are evaluated:
    OR – A point is classified as positive if any threshold is met. Use for broader multi-element targeting.
    AND – A point is classified as positive only if all thresholds are met. Use when targeting a specific element combination.
Review Target Element Breakdown
- This table is auto-populated based on the selected thresholds.
  - It displays the number/percentage of mineralized vs. unmineralized points
Complete Step
- Click Generate Learning Data to save and continue.
  - This step takes approximately 2 minutes to complete.

Considerations When Adjusting the Threshold

The recommended threshold can be a strong starting point to help reduce guesswork. However, there are many cases where strategic adjustments may be appropriate.

Here are two examples of situations where you might want to adjust the threshold cutoff:

You are working with an economically (e.g., internal reporting or market-driven) defined cutoff grade
The AOI represents only part of a larger mineral system. In this case, you may want a threshold that aligns with other deposits in the district, regional mineralization style, or a broader geological context. Consistency across targets may be more important than the localized clustering logic.

When adjusting manually, aim for approximately 10–20% mineralized examples, as displayed in the Target Element Breakdown table. Too few positives limit the model’s ability to learn meaningful patterns, while extreme imbalance can reduce performance.