Skip to content

View a model (experiment) summary

A model summary contains an array of summary metrics about a Driverless AI (DAI) experiment in your established DAI connection (e.g., important features, scorer, validation score, etc.).

Instructions

To access a model summary, consider the following instructions:

  1. In the H2O Model Validation navigation menu, click Experiments.

    Note

    • On the Experiments card, in particular, in the experiments table you can view a model summary for all your Driverless AI (DAI) experiments (models) in your established DAI connection. To learn more, see Experiments table columns.

    • The experiment summaries in the Experiments card are for the experiments found in the established DAI connection. To learn more, see Create a connection.

  2. In the experiments table, select the model summary you want to view.

  3. Click View.

    Note

    When viewing an experiment, the following chart and graph appear as part of the model summary metrics:

Experiments table columns

Column name
 Description
Name Experiment name.
Model summary State of model summary. See Model summary states to learn about all the different states for the model summary column.
Scorer Experiment scorer.
Val score Experiment validation score value.
Test score Experiment test score value.
Train dataset name Name of the experiment train dataset.
Test dataset name Name of the experiment test dataset.
Adversarial Similarity The number of Adversarial Similarity tests complete and scheduled to run.
Backtesting The number of Backtesting tests complete and scheduled to run.
Drift Detection The number of Drift Detection tests complete and scheduled to run.
Size Dependency The number of Size Dependency tests complete and scheduled to run.

Model summary states

As follows are the different types of states a model summary can be in:

  • NotCreated

    H2O Model Validation has not created the model summary.

  • Created

    H2O Model Validation created the model summary.

  • Running

    H2O Model Validation is currently running the model summary.

  • Done

    H2O Model Validation has completed the model summary.

  • Deleted

    H2O Model Validation deleted the model summary.

  • Error

    An error occurred during the model summary.

  • Timeout

    There was not enough time to complete the model summary.

Metrics

Graph: Tuning process

The tuning process graph displays the iteration data of the experiment.

  • X-axis: Iteration (since Driverless AI is an AutoML tool, it has several iterations before the final modelling pipeline.)
  • Y-axis: Depends on the scorer of the experiment (e.g., RMSE, AUC)
Graph: Tuning process

Chart: Feature importance

The feature importance chart displays the top features of the experiment.

  • X-axis: Gain value (the feature's importance to the model)
  • Y-axis: Feature name
Chart: Feature importance