Question

James.Thomas100 on Thu, 21 Apr 2016 11:52:43


Hi,

I'm using the advice given here to create a ROC curve in python: https://social.msdn.microsoft.com/Forums/azure/en-US/5d22a159-fade-49ac-b7d3-a0e7a1bc895c/there-is-no-score-column-in-scored-dataset?forum=MachineLearning

This works fine, when I want to view the curve in the ML studio, but how should I adjust the code so that I can export in the Dataframe and save as csv - in a format that I can then import and decode elsewhere? (i.e. I need to deploy this as a web service.)

Apologies if this is obvious - I've experimented with a couple of options, and drawn a blank so far.

Thanks, James.


Sponsored



Replies

Amy Kate Nicholson on Thu, 21 Apr 2016 13:35:58


Hi James,

Just want to check a few details on this one. Are you using an 'Execute Python Script' Module? containing python code to create an ROC curve graph? and you can view this graph in the Python Device output?

Can you clarify what you mean by decode? I was initially thinking you can open the output or input dataset in a Python Jupyter Notebook and do more data exploration in there, separate from your experiment. And therefore you could create a predictive web service from your model in the experiment space. You can also just right click save as the graph if that is the part you want to keep for a document for example

Maybe a screenshot of your experiment setup could be useful here?

Thanks,

Amy

neerajkh_MSFT on Fri, 29 Apr 2016 02:39:53


You can use web service output module and attach to PNG port of python (see picture below)

James.Thomas100 on Thu, 28 Jul 2016 08:03:07


Thanks (and apologies for the slow response) - this looks really helpful, and good to have two outputs from the python script!