Overview

In this article, we explain what triggers DORA's “Select Input Features Failed” error, what the checks are looking for, and what you can do to fix the problem before re-running your experiment.

What is the “Select Input Features Failed” Error?

When you click Run Prediction in Step 4, DORA automatically validates your input features before proceeding. You may see this error message, or something similar:

For the prediction to run, DORA requires a minimum of 100 valid Learning Point cells within the shared raster overlap area.

💡Note: The point counts shown in the Target Element Breakdown table in Step 2: Set Up Learning Points are not the same as cells. DORA converts points into raster cells before running the prediction, and many points can collapse into a single cell depending on raster resolution. A large point count in Step 2 does not guarantee the 100-cell minimum will be met.

The prediction cannot proceed if any of the following conditions are met:

Too few Learning Points fall within the shared raster overlap area. This occurs when no Learning Points are present, or when more than 90% of your Learning Points fall outside it.
The Learning Points within the overlap area do not include enough positive and negative examples. This occurs when one type is missing entirely, or when fewer than two of either type remain.
Any selected raster does not overlap spatially with every other selected raster. DORA compares all rasters against each other, so even one non-overlapping raster will trigger this error.

Rasters do not overlap with other rasters

Review the options below to resolve the issue before re-running the prediction.

How to Fix the Error

If you receive this error, there are three approaches to consider:

Option 1: Review and remove non-overlapping rasters

Check which rasters do not share coverage with the rest of your selection and remove them. Reducing your raster set to only those that overlap spatially is often the fastest fix. If your rasters cover different spatial zones, it may still be possible to run the prediction using a single raster from one zone, provided it has sufficient Learning Points within its coverage area.

If you are unsure which raster is causing the overlap failure, reach out to your DORA contact for assistance.

Option 2: Adjust your AOI

Resize or reposition your AOI to focus on the area where your rasters and Learning Points overlap. A smaller, well-covered AOI is more effective than a large one with patchy raster coverage.

Option 3: Increase the raster resolution manually

If your rasters are coarse, DORA's automated resolution setting may not generate enough Learning Point cells to meet the 100-cell minimum. In this case, manually setting the height and width resolution from the Advanced Settings option in Step 4: Select Input Features can increase the number of cells produced.

If you are unsure what resolution values to use, reach out to your DORA contact for guidance.