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

Confidence interval Decision tree

MaPerez
7 - Meteor

Greetings,

 

I have used a desicion tree (regresion) to estimate some transactions. I need to calculate the confidence interval, but when I checked the option "confidence interval" in the score tool, the interval doesn't show.

 

Can anyone help me?

 

Thanks!!!

4 REPLIES 4
BenMoss
ACE Emeritus
ACE Emeritus

If you take a look at the documentation for the score tool there is a note about the fact that you can only use the 'confidence interval' option on 'non-regularized' linear regression models only.

My guess is that the 'forest model' does not meet this criteria, indeed according to this post (https://datascience.stackexchange.com/questions/6787/is-decision-tree-algorithm-a-linear-or-nonlinea...) a forest model is not a linear model at all.

 

On a side note, when you shared your work you sent both the .yxmd file and the .yxdb file to allow us to run the workflow correctly; it's actually possible to create a .yxzp file which contains both of these and can be shared as a single upload, it also makes all paths referential meaning the user simply has to open it rather than reconfigure the inputs and outputs. You can create a .yxzp file by going to options > Export Workflow.

 

Ben

Charity_K_Wilson
10 - Fireball

As Ben said, Confidence Intervals are not a part of the Decision Tree, Forest Model, nor the Neural Network.  I took your data and ran it thru a Linear Regression, and a Count Regression.  Both of these types of regression do allow for a confidence interval.  You have the standard error, and the t-value, and can calculate the confidence band around each.  

MaPerez
7 - Meteor

Thanks Ben for your response!!!

 

MaPerez
7 - Meteor

Thanks Charity for your response!!! It was very helpfull!!

Labels