Uploaded image for project: 'Public H2O 3'
  1. Public H2O 3
  2. PUBDEV-3908

Expose metalearner & base model keys in R/Py StackedEnsemble object

    Details

    • Type: Improvement
    • Status: In Progress
    • Priority: Major
    • Resolution: Unresolved
    • Affects Version/s: None
    • Fix Version/s: None
    • Component/s: Python, R, StackedEnsemble
    • Labels:
      None
    • CustomerVisible:
      No

      Description

      Currently, a "Stacked Ensemble" model object only contains the info/metadata that a normal "H2OModel" has. We should add the ability to access the name/ID of the metalearning model (at the very least) and also the names of the constituent base models.

      Should there be a superclass of "H2OModel" that has these extra attributes?

      Right now the `my_ensemble@model` object contains the following:

      > my_ensemble@model
      $cross_validation_models
      NULL
      
      $cross_validation_predictions
      NULL
      
      $cross_validation_holdout_predictions_frame_id
      NULL
      
      $cross_validation_fold_assignment_frame_id
      NULL
      
      $model_summary
      NULL
      
      $scoring_history
      NULL
      
      $training_metrics
      H2ORegressionMetrics: stackedensemble
      ** Reported on training data. **
      
      MSE:  0.1558116
      RMSE:  0.3947297
      MAE:  0.3468965
      RMSLE:  0.2788961
      Mean Residual Deviance :  0.1558116
      
      
      $validation_metrics
      H2ORegressionMetrics: stackedensemble
      ** Reported on validation data. **
      
      MSE:  
      RMSE:  
      MAE:  
      RMSLE:  
      Mean Residual Deviance :  
      
      
      $cross_validation_metrics
      H2ORegressionMetrics: stackedensemble
      ** Reported on cross-validation data. **
      
      MSE:  
      RMSE:  
      MAE:  
      RMSLE:  
      Mean Residual Deviance :  
      
      
      $cross_validation_metrics_summary
      NULL
      
      $status
      NULL
      
      $start_time
      [1] 1.484704e+12
      
      $end_time
      [1] 1.484704e+12
      
      $run_time
      [1] 211
      
      $help
      $help$cross_validation_metrics_summary
      [1] "Cross-validation model metrics summary"
      
      $help$run_time
      [1] "Runtime in milliseconds"
      
      $help$status
      [1] "Job status"
      
      $help$domains
      [1] "Domains for categorical columns"
      
      $help$model_category
      [1] "Category of the model (e.g., Binomial)"
      
      $help$`__meta`
      [1] "Metadata on this schema instance, to make it self-describing."
      
      $help$model_summary
      [1] "Model summary"
      
      $help$scoring_history
      [1] "Scoring history"
      
      $help$help
      [1] "Help information for output fields"
      
      $help$end_time
      [1] "End time in milliseconds"
      
      $help$names
      [1] "Column names"
      
      $help$cross_validation_fold_assignment_frame_id
      [1] "Cross-validation fold assignment (each row is assigned to one holdout fold)"
      
      $help$start_time
      [1] "Start time in milliseconds"
      
      $help$training_metrics
      [1] "Training data model metrics"
      
      $help$cross_validation_models
      [1] "Cross-validation models (model ids)"
      
      $help$cross_validation_metrics
      [1] "Cross-validation model metrics"
      
      $help$cross_validation_predictions
      [1] "Cross-validation predictions, one per cv model (deprecated, use cross_validation_holdout_predictions_frame_id instead)"
      
      $help$cross_validation_holdout_predictions_frame_id
      [1] "Cross-validation holdout predictions (full out-of-sample predictions on training data)"
      
      $help$validation_metrics
      [1] "Validation data model metrics"
      

        Attachments

          Activity

            People

            • Assignee:
              rpeck Raymond Peck
              Reporter:
              erin Erin LeDell
            • Votes:
              0 Vote for this issue
              Watchers:
              1 Start watching this issue

              Dates

              • Created:
                Updated:

                Zendesk Support