View Logistic Regression Model Weights in Machine Learning Studio Enterprise

Karl Märka 1

In ML Studio Classic you can see the logistic regression model weights (intercept and coefficients) when you click on the output from "Train model" module. This feature however is not available in the new ML Studio Enterprise. I only get the Trained_model blob in blob storage. Seeing the weights is a must have when building credit scoring models and this current deficiency is what's keeping us from switching to ML Studio Enterprise.

Is there a plan to re-introduce this feature from the ML Studio Classic? Alternatively is there a way I can serialize the data.ilearner file and get the model weights from there?

I know I could use the "Train Python model" module instead, but the "Evaluate Model" module fails with an unspecified error when I use that.

romungi-MSFT 41,961 Reputation points Microsoft Employee

2020-06-08T10:57:57.667+00:00

@Karl Märka Thanks for the question. We are checking this internally to understand if the same feature will be added in the designer studio.
With respect to the unspecified error in your current experiment could you please submit the details of the run from the Azure ML portal from the top right hand corner by clicking the smiley face? This will let our service team know about the error and they have an option to contact you regarding this if you select the option to contact you while submitting this feedback.
romungi-MSFT 41,961 Reputation points Microsoft Employee

2020-06-16T08:09:19.887+00:00

@Karl Märka Is it possible to provide details of the difference in classic studio and designer in an email to AzCommunity[at]microsoft[dot]com so our product team can work with you to understand this scenario and advise?