[Python] why the performance results of compare_models and plot_model for a setup with 10 fold...

In the 10 fold cv session, there is no extral "test_data" for setup, so the data were splitted into traning data and testing data per fold.

Is the result of compare_models from an average of performance of training + testing data with 10 fold? The result of plot_model is from an average of performance of "hold-out" dataset as shown in the manual. Is the "hold-out" datasetthe testing data per fold here and the results of plot_model are averaged with 10 folds of testing data?

As the performance results of compare_models and plot_model are different. Thank you.

Continue reading...