Interpreting cross-validation results
Recall that the result variable we stored after the experiment.Execute call is a Cross
ValidationExperimentResult<RegressionMetrics> instance.
If we end a cell with that object in Polyglot Notebooks, we see a graphical representation of the results, as shown in Figure 8.7:
Figure 8.7 – Our experiment results, including RunDetails and BestRun
Much of this is similar to what we saw with a standard train/test split in the previous chapter: we still have a RunDetails property listing all models evaluated and we still have a BestRun property containing the most effective model according to the evaluation metric we specified in our settings object.
What’s different here is that each run details contain not just a single model and set of metrics, but one model and associated metrics for every cross-validation fold that was evaluated.
Let’s illustrate this by displaying the MeanAbsoluteError...
