Using Your Predictions

Once a model wins the search, ML Studio puts it to work automatically. The prediction becomes part of your dataset's enrichments, so every case gets predicted values that refresh whenever the data is reloaded. This page covers where those values appear, how to read them, and how to manage your saved predictions.

Where the predicted values appear

A finished prediction adds new columns to your cases. You'll find them in Case Explorer (and anywhere else case attributes are used - filters, dashboards, and exports) after you reload the dataset.

Case Explorer showing predicted columns alongside regular case attributes

The predicted columns are named after your target:

  • Predicted [target] - the predicted value for each case. For Binary and Classification this is the predicted label (for example a class like Rehab or a 0 / 1 flag); for Regression it's the predicted number (for example 4.75 days).
  • [target] confidence - for Binary and Classification, how confident the model is in that prediction, shown as a percentage. (Regression predictions report the number directly and don't add a separate confidence column.)

Because each prediction is just a set of case attributes, you can build more than one on the same dataset. In the example above, a case carries a predicted length of stay, a predicted discharge disposition, and a predicted readmission flag with its confidence - all at once.

Reading the values

  • Sort or filter on a Predicted [target] column to find the cases that matter - the ones predicted to fail, to be late, or to fall into a category you care about.
  • Use the confidence to focus on the predictions the model is most sure about, or to review the borderline ones.
  • Combine with your other attributes to act - route high-risk cases, prioritize work, or trigger an alert.

Keeping predictions current

Predicted values are recalculated every time the dataset is reloaded, so they stay in step with your latest data. If you add new cases or new activity, reload the dataset to refresh the predictions.

Managing your predictions

Every prediction you build is saved with the project and listed under Your predictions on the ML Studio landing screen. Each entry shows the target, the prediction type, when it was trained, and the winning model's score. From there you can:

  • Open the training report - revisit the full scorecard, metrics, drivers, and leaderboard for that prediction at any time.
  • Retrain - rebuild the model on the current data. Use this after your process or data has changed, so the prediction learns from the most recent cases.

Retraining searches the models again and updates the same predicted columns in place, so anything already built on top of them keeps working.

Tips

  • Start with Binary. A yes / no prediction is the quickest way to get a useful, easy-to-read result.
  • Enrich first. The more meaningful enrichments your dataset has, the more the model has to learn from - and the better the prediction.
  • Retrain periodically. Processes drift over time; an occasional retrain keeps predictions accurate.

Providing feedback

ML Studio is an alpha feature and your feedback shapes it:

  • Email: support@mindzie.com
  • Subject: include "Alpha Feedback: ML Studio"
  • Include: what you were trying to do, what happened, and what you expected.