José Roberto Pereira · Dec 27, 2020 2m read

Using IntegratedML to create a cube dimension

Hi guys.

That time I'll show you a way to use a machine learning model implemented using IntegradeML as source for an IRIS analytics cube dimension.

Creating the ML model:

Execute this SQL in order to create a new ML model using IntegratedML SQL extension:

CREATE MODEL AppointmentsPredection 
FROM dc_myapp_model.MedicalAppointments

Now, you can train you model:

TRAIN MODEL AppointmentsPredection FROM dc_myapp_model.MedicalAppointmentsTrain

Finally, you can see how good your model is:

VALIDATE MODEL AppointmentsPredection FROM dc_myapp_model.MedicalAppointmentsTest

Creating the cube dimension:

The trick is to use an expression as input for the dimesion. This is done by setting the Expression property of the dimension:

The method expression code uses the PREDICT function of IntegratedML to classifiy an new entry using the ML model created early:

ClassMethod PredictMedicalAppointmentClass(pID As %String) As %String
  Set modelClass = ##class(dc.myapp.model.MedicalAppointments).ShowGetStored(pID)
  If modelClass '= "" {
    Return ""
  &SQL(SELECT PREDICT(AppointmentsPredection) INTO :modelClass FROM dc_myapp_model.MedicalAppointments WHERE ID = :pID)
  return $Case(modelClass, 0:"No show", 1:"Show", :"")

So, with this aproach you can create a dimension which will show to users a prediction about future appointments.

Using your new cube in a notebook:

As now you has a cube with dimensions that can show past appointments as do predictions about future appointments, you can explore this information in a notebook.

For example, you can create a pivot table showing the history of past appointments:

And ,you can also show the forecast for future appointments using the dimension which uses an IntegratedML prediction model:

You can access this example in my aplication by seaching a notebook called "medical-appointments".

1 0 0 69
Log in or sign up to continue