Skip to main content

Task 5: Summary tab

Overview​

In the MLI (Machine Learning Interpretability) report of an experiment, the SUMMARY tab provides a concise overview of interpretability findings, highlighting key features and evaluating the model's overall performance. The content in the SUMMARY tab may vary depending on the experiment type (regular or time-series) and the interpretability methods used. Typically, it presents the most important variables (features) of the H2O Driverless AI model, ranked by importance in descending order, and indicates whether these features are original or transformed. Additionally, in the SUMMARY tab, when the MLI report originates from an H2O Driverless AI model, it includes a table summarizing the H2O Driverless AI model.

  1. Click the SUMMARY tab.
    note

    To learn more about the Summary tab, see Summary tab.

Validation score​

caution

From this point, your model's metrics might differ from those discussed here.

In the SUMMARY tab, we can observe that the H2O Driverless AI model received a validation (R2) score of 0.8474 for the time series model.

An R2 score of 0.8474 with a small standard deviation (± 0.002624016) in the context of a time series model in H2O Driverless AI indicates a strong performance. An R2 value of 0.8474 closer to 1 suggests that the model can explain approximately 84.2% of the variance in the target variable. This is generally considered a good fit, indicating that the model effectively captures the underlying patterns in the data.

The low standard deviation (± 0.002624016) indicates that the model's performance is consistent across different validation sets. This consistency is important for reliability and robustness.

Values Validation score

note

Whether the validation score is considered "good" depends on the context and domain. In other words, a validation score can be considered good if it meets the specific needs of your application and outperforms baseline models.

Top 4 most important variables (features)​

For the time series model, in the SUMMARY tab, the top 4 most important variables (features) for the model, all originating from the original features (variables), are as follows:

  • Dept (Original): This feature represents the column that serves as the identifier for a department within a store.
  • Date (Original): This feature represents the column that identifies the end date of a week (in other words, the end date of the week's sales record).
  • Store (Original): This feature represents the column that serves as the identifier for a store.
  • IsHoliday (Original): This feature represents the column indicating whether a week contains a major holiday.

Top features

Feedback