Friday 20 April 2018

Run and deploy in Azure ML Studio - Part 2

  1. Run the Model.

      
  1. After completing the model run. You can visualize the data by right clicking on the Score model and navigating to the Visualize as below.  


  1. After visualizing the data, you can see the Scored labels(output). This is the predicted value that comes out of the tested dataset. It will lie between -3 to 3 
  1. Close the visualize popup and Select Setup Web Services and click onUpdate Predictive Experiment 
  1. This will create the predictive experiment canvas as a tab next to training experiment. Then selectSelect Column in Datasetand then click onLaunch Column Selector”.   
  1. Remove the symboling (safety factor) feature so it will not be part of input from the external api call. And Click save 
  1. Run  the model. 
  1. Search forSelect columns in Datasetmodule and add the module in the predictive experiment canvas and then connect them to score model and Web Service output as below 
  1. Remove all columns except Scored Labels feature on clicking on theLaunch Column Selectoras below. 


  1. Run the model again 
  1. Then SelectDeploy Web Service 
  1. Once the Deploy web service is executed, it navigates to predictive experiment dashboard, where you can test the web service. Easiest way to test the web service is to click on theTestlink show below. 
  1. After clicking the test link it moves to “Microsoft Azure Machine Learning Web Services”. For easier testing enable the Sample data feature. 
  1. On enabling you will see the data being filled automatically and then click on Test. 
  1. You can see the predicted result on how safe the car is. 

Build Bot using LUIS