Prediction Fusion

Fuses multiple prediction confidences into one, combined prediction, using the selected fusion method. The selected fusion method will be applied to all confidence values (resulting from multiple predictions) of the same class. After the fusion method was applied, the resulting combined prediction confidences will be normalized to add up to one, such that they can be interpreted as probabilities.

Note: Missing values will be skipped.

Options

Method

The fusion method.

• Maximum: Maximum of all prediction confidences
• Mean: Mean of all prediction confidences
• Median: Median of all prediction confidences
• Minimum: Minimum of all prediction confidences

Classes column

String column from which to extract the classes from it's domain.

Add classes from column domain

Add classes from the domain of the selected string column.

Classes

The classes that were predicted and for which confidences are available. In case of a tie the higher class (in the list) wins.

• Class name: The name of the class.

Class confidences

Each row represents the predicted confidences from one classifier. The fusion method is applied to all confidence values of the same class.

• Weight: Multiplies the prediction confidences as if the prediction would have been added n times. This enables a prediction to have more influence on the outcome than another. (Affects mean and median but not minimum or maximum.)
• Confidence columns: Select the confidence column for each class. The columns must be compatible with double values.
  • Column: Select the confidence column for this class.

Views

This node has no views

Workflows

• 01_AnalyticsKNIME Hub
• 01_Combining_Classifiers_using_Prediction_FusionKNIME Hub
• 04_Cross-Platform_Ensemble_ModelKNIME Hub
• StromverträgeKNIME Hub
• Stromverträge_2KNIME Hub

Links

• Combine results from two classification modelsby qqilihq on 2020-01-17

  forum.knime.com/p/79091

Developers