There are several analyses and reports that can be run against a model. Global Feature Importance will show which features or columns contribute the most, globally, to predictions by the model. Batch Query Analytics will examine the output of a model via a batch query and display a cumulative performance analysis.
Global Feature Importance examines the model, directly, and lists features or columns by relative weight in determining predictions. This is a static analysis of expected importance. Due to ML Studio’s explainability, each prediction will have its own relative weighted factor list. Therefore the output of this report should be used as a guide about model behavior in general.
Batch Query Analytics looks at model performance with respect to a specific batch query. Specifically, it requires a batch query that also contains ground truth values for the predicted class. This report will compare the predictions against the ground truth and show classification rates and counts by probability levels, and corresponding cumulative values.