Parsing your AutoML results via AMLS and the AML SDK

When the experiment run is completed, we are able to extract valuable information from the AutoML experiment run by leveraging the AML Python SDK:

1. For an AutoML experiment run, each run is executed as a child run of the AutoML experiment run. This means that we can get the best run for the experiment run by looking at the child runs for a given run as shown in Figure 5.29:

Figure 5.29 – Retrieving the best child run for an AutoML experiment run

2. As we can view the information programmatically, we are also able to retrieve the best model through AMLS. In the studio, by clicking on the automl-classification-titanic experiment, we can see the run created for the experiment.
3. Clicking on the run hyperlink brings us to the details of the experiment run. If we move over to the Models tab, we can see the models by their values for the primary metric, and an explanation provided for...