Variable importance in Linear regression

Question

Hi,

Am running a Linear regression model in Azure ML studio having multiple features(both numeric and categorical).

Is there a way to get the important features among all the given input features?

From the train model I see weights assigned to each of the variable. Does Azure ML regression normalize the variables before running the model. If so can we assume this weight for the important features ?

Can we export this weights to csv file ?

Answer 1

Hi, thanks for reaching out. AML has two modules for feature selection, please check out Filter Based Feature Selection and Permutation Feature Importance. Regarding the weights, are you referring to the output from regression model? If so, here's a description of the output parameters. To export model results, you can use the export data module to export results and data to any of the supported cloud services. Hope this helps.