Feature Selection Visualizations

The following tables and options are available for Feature Selection visualizations.

Model Information table

Displays input settings used to control how features are screened based on missing data or little or no variability, settings to control how features are classified as Important, Marginal, or Unimportant, and whether fields of each level of importance are to be retained. For categorical targets, shows the test statistic used to generate p values to classify categorical targets in terms of importance. This can be a Pearson or likelihood-ratio chi-square test, Cramer's V, or Lambda.

Feature Importance table

This table shows a ranked list of features in order of descending importance for predicting the target. It includes the measurement level for each feature, an assignment of its importance as Important, Marginal, or Unimportant, and a true or false flag column that indicates which fields would be retained according to the importance cutoff specified when running the algorithm. The final column, labeled Value, shows 1 – p, where p is the significance value for a test of association between the feature and the target.

Next steps

Like your visualization? Why not deploy it? For more information, see Deploy a model.