Models ensembling

You can ensemble prediction models in DSS using various ensembling methods.

In the analysis results page, select several models, and click Actions > Create ensemble model.

Limitations

  • Only models trained with the same preprocessing settings can be ensembled.

  • It is not possible to ensemble models with different time ordering or weight parameters.

  • It is not possible to ensemble models trained using K-Fold cross-test.

  • It is not possible to ensemble models trained with vector or image features.

  • It is not possible to ensemble partitioned models.

  • It is not possible to ensemble computer vision models.

  • After deployment as a saved model, it is not possible to retrain an ensemble model on the automation node.

  • After deployment as a saved model, it is not possible to retrain an ensemble model if the original models have been removed from the analysis screen.