Use Predicted Value from One Workflow in a Separate Workflow

[jlrouquette]

I'm using the support vector machine tool to make a prediction on some categorical data.  I would like to input that prediction into another workflow I'm working with.  When I use predictive alternatives like JMP (a SAS product) I can get a prediction expression where I apply the predictor variables to an equation to get the predicted value. 

When it comes to doing something with the predicted value here, though, I'm struggling with how to do anything other than see the value.

Any help would be much appreciated.

-Luke    

[DavidP]

A couple of options come to mind.

If you turn them both into apps you can chain the apps so that completion of the 1st triggers the running of the 2nd. In this case you can write the output of the 1st app to a temp alteryx db file and load this file as the input to the 2nd app.

You could also turn your 2nd workflow into a macro with a macro input that is inserted into your 1st workflow.

[WilliamR]

Hi,

if I understand well, you want to apply your model to other dataset right ?

If so, you can store your model in a .yxdb file using an output data connected to the O anchor of the model, then on the other workflow you import your model with the input data tool and score your data through the score tool.

Regards,

[WilliamR]

Here is what it looks like:

Workflow 1: training the model

Workflow 2: scoring data

[jlrouquette]

I appreciate the responses from both of you.  Its possible both could address the need, maybe I'm just still too green with Alteryx.  There is a bit of an order of operations issues that I should elaborate on I think. Some of this process might be convoluting my interpretation of your suggestions.

I have a workflow that is creating a dataset.  Latitude and longitude, per observation, are part of that dataset.  We have a unique geographic cluster that I have to assign every lat and long, i.e. what I'm using the model for.  

Pull lat & long (with full workflow) --> predict geographicclusterr --> dump back into workflow --> output complete dataset

Does this extra context change any of how you would suggest addressing the issue?

[WilliamR · Accepted solution]

Luke,

please find in attachement a workflow with auto-selection based on a statistical metric (MAPE). The idea behind is that the model code can be processed like any other data and used later in the worklow with a score tool.

Hope this help.

[jlrouquette]

Really appreciate your help with this, William!