Community Spring Cleaning week is here! Join your fellow Maveryx in digging through your old posts and marking comments on them as solved. Learn more here!

Alteryx Designer Desktop Discussions

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

Reusing Regression output to score different data sets

Ashish
8 - Asteroid

Hi,

 

I want to set up a Linear/(or any other) regression model work flow in such a way that I can use the output of the model to score different sets of data without running the model. 

Basically my intent is to train the model today and then run it every day for few weeks/days (on incoming new data) to see the results and then train again after couple of weeks and test for few days. 

 

Also is there a way to dyanamically controll if I want to re run the model or not ? 

I want to do it because it wil be important feasibility factor for scalability.  

 

 

Score without running the model

2 REPLIES 2
RodL
Alteryx Alumni (Retired)

Ashish, 

First of all, I want to make sure you realize that you can save the "model object" that is created in most of the Predictive tools. All you need to do is put an Output tool after the 'O' side of the tool and save to a YXDB file format. Then you can use an Input tool and point to that model and then use that for scoring on any new data. (The caveat on that, of course, is that the data needs to be in the same format as what was used to build the model that you saved. That includes variables themselves, as well as the values within a predictor variable (for example, if you have new data that has a categorical value not included in the original model, it won't know how to use that in the scoring process). 

 

With that said, what I would do is convert the process into an app where you have radio buttons to either use the existing model or create a new one. That would be tied to actions that disable one of two Tool Containers which would either use the existing model, or create a new one. You can add a Union tool to both so that one model will be used with the data set to be scored. 

The user of the app would just need to be careful in understanding what the ramifications were for their choice.

 

Conditional Modeling.png

 

I have attached a sample for you to use as an example to possibly build on.

If you'd prefer not to have it as an app, but just run it in Designer, you can eliminate the questions and just manually disable whichever Tool Container you don't want at the time.

Hope this helps.

Rod

Ashish
8 - Asteroid

Thanks a lot! This is exactly what I was looking for!

Labels