AutoML: How to diagnose the selected model?

[WSDATA]

Tinkering with the AutoML learning tool and comparing it to the Assisted modeling path.

In the assisted modeling flow, you can compare models, see feature importance, etc.

Can anyone help me understand how can I learn more and diagnose the model coming out of AutoML?

[WSDATA]

OK, I can see from the workflow log what the selected model is coming out of AutoML.

I have also been able to hook the AutoML up to a "predict" tool to apply the model to new data that the model has not seen.

I am still looking to learn more about the model that AutoML selected.

Other than the "predict" tool, are there any other tools that can accept the M (Model) output from AutoML?

Any help is appreciated!

[ArtApa]

Hi @WSDATA - The AutoML tool automatically selects the best machine learning method based on the target you've selected. Read more in the documentation: https://help.alteryx.com/20214/designer/automl 

[WSDATA]

Art,

Thanks for the feedback.  I can see that AutoML selects the best approach.  I can also get that model out to the predict tool to score new data.

However, I can not figure out how to get detailed information about the model that it selected.

For example, which features are more or less important?

I am guessing that they are using EvalML under the hood.  If you read the docs on EvalML, you can see a lot of diagnostic information is available.

I am wondering if it is worth skipping the ease of use of AutoML and just call EvalML from a Python tool.

I guess I was hoping I could feed the AutoML model into another tool that would allow me to look at the details, but I have not found that.

Any help/direction is appreciated!

[WSDATA]

Does anyone out there have more experience with AutoML that they can share?  Any help would be fantastic.