Alteryx Designer Desktop Discussions

Find answers, ask questions, and share expertise about Alteryx Designer Desktop and Intelligence Suite.
SOLVED

How to retrain Alteryx predictive model?

mohamedhussen
5 - Atom

Hello, 

I trained the Alteryx Boosted Model on my data before and save the model to yxdb file. now I'm getting new data and need to fine-tune the model and retrain it to the new data so how can I fine-tune this model in Alteryx?

 

Also, Is there a way to save this model and use it inside my python code to make predictions?

 

Thanks

4 REPLIES 4
fmvizcaino
17 - Castor
17 - Castor

Hi @mohamedhussen ,

 

The model customization can be found in the second tab of the boosted model, you can't fine tune the model after saving it to a yxdb file, you need to do it before.

fmvizcaino_0-1609959000144.png

 

You can create a workflow with the boosted model and use alongside a python tool, but it is not possible to export the model and use inside python.

 

Best,

Fernando Vizcaino

 

CharlieS
17 - Castor
17 - Castor

Hi @mohamedhussen 

 

Each time you run the workflow, the model is trained based on the input data and the workflow configuration/setting. If either the data or the settings change (like @fmvizcaino  outlined above), you must save off the model object .yxdb to retain each version of the model. 

 

If you want to dissect the information within each model object, there's a bit of R you can use to view/extract the values of interest. Here's a post of mine outlining that process:

https://community.alteryx.com/t5/Alteryx-Designer-Discussions/Extract-the-logic-behind-a-pedictive-m... 

 

However, it will likely be the case that a boosted model has too many attributes for this decomposition method to be useful. It's worth a shot, but it typically more useful on other model types. 

mohamedhussen
5 - Atom

Hi @fmvizcaino 

Thanks for your reply

So to fine-tune my model I need to train it with the old data and new data together, right? and there is no way to apply transfer learning on the models that I trained before with Alteryx predictive models? 

fmvizcaino
17 - Castor
17 - Castor

Hi @mohamedhussen ,

 

Yes, you always need to use old + new data to recreate your model. 

Unfortunately, there is no way of doing transfer learning.

 

If you have the intelligence suite package (paid add-on), you can generate the python code from your model and then use the python tool to do transfer learning, I think. But the assisted modelling, which gives you the ability to do it doesn't have all the visual models Alteryx have yet.

https://help.alteryx.com/pt-br/current/designer/alteryx-intelligence-suite

 

Best,

Fernando Vizcaino

Labels